PR libstdc++/87308 adjust regex used in std::any pretty printer
[official-gcc.git] / gcc / ChangeLog
blob7a2ee245c39d0aee96231bc6699f68dd225f601d
1 2018-11-23  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/88149
4         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
5         the case where there are two different def types for the
6         same operand at different operand position in the same stmt.
8 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
10         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
11         gen_atomic_compare_swap_1.
12         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
13         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
14         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
15         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
16         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
17         checks.
18         * config/arm/arm.md (neon_vext<mode>),
19         (neon_vrev64<mode>, neon_vrev32<mode>),
20         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
21         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
22         before the pattern name.
23         * config/arm/sync.md:
24         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
25         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
27 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
29         PR tree-optimization/86614
30         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
31         if TREE_NO_WARNING is set on ref.ref.
33 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
34             Alan Coopersmith  <alan.coopersmith@oracle.com>
36         PR c/53608
37         * doc/extend.texi (Designated Inits): Clarify handling of multiple
38         initializers for unions.
40 2018-11-22  Jan Hubicka  <jh@suse.cz>
42         PR lto/88142
43         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
44         parameters; do not warn here.
45         (odr_subtypes_equivalent_p): Likewise.
46         (warn_odr): Fix typo.
47         (warn_types_mismatch): Do not output confused warnings on integer types.
48         (odr_types_equivalent_p): Update.
50 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
52         * config/i386/i386.c (ix86_option_override_internal,
53         ix86_can_inline_p, classify_argument, construct_container,
54         ix86_expand_prologue, ix86_expand_split_stack_prologue,
55         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
56         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
57         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
58         Formatting fixes.
60         * config/i386/i386.c (ix86_option_override_internal): For
61         stack_protector_guard related options, use opts_set->x_ instead
62         of global_options_set. and prefix options with opts->x_ .  Move
63         defaults for offset and reg into else block.
65 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
67         * config/i386/i386.c (ix86_check_avx_upper_register):
68         Return true for all SSE registers with mode bitsize > 128.
70 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
72         PR rtl-optimization/87718
73         * ira-costs.c: Remove trailing white-spaces.
74         (record_operand_costs): Add a special treatment for moves
75         involving a hard register.
77 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
79         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
80         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
82 2018-11-22  Martin Liska  <mliska@suse.cz>
84         * common/config/i386/i386-common.c (processor_names): Add
85         static assert and add missing "znver2".
86         (ix86_get_valid_option_values): Add checking assert for null
87         values and add "native" value if feasible.
88         * config/i386/i386.h: Do not declare size of processor_names.
89         * common/config/i386/i386-common.c:
90         * config/i386/i386.c: Add static assert for size
91         of processor_cost_table.
93 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
95         * target-insns.def (stack_protect_combined_set): Define new standard
96         pattern name.
97         (stack_protect_combined_test): Likewise.
98         * cfgexpand.c (stack_protect_prologue): Try new
99         stack_protect_combined_set pattern first.
100         * function.c (stack_protect_epilogue): Try new
101         stack_protect_combined_test pattern first.
102         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
103         parameters to control which register to use as PIC register and force
104         reloading PIC register respectively.  Insert in the stream of insns if
105         possible.
106         (legitimize_pic_address): Expose above new parameters in prototype and
107         adapt recursive calls accordingly.  Use pic_reg if non null instead of
108         cached one.
109         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
110         (arm_legitimize_address): Adapt to new legitimize_pic_address
111         prototype.
112         (thumb_legitimize_address): Likewise.
113         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
114         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
115         (thumb1_expand_prologue): Likewise.
116         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
117         change.
118         (arm_load_pic_register): Likewise.
119         * config/arm/predicated.md (guard_addr_operand): New predicate.
120         (guard_operand): New predicate.
121         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
122         prototype change.
123         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
124         prototype change.
125         (stack_protect_combined_set): New expander..
126         (stack_protect_combined_set_insn): New insn_and_split pattern.
127         (stack_protect_set_insn): New insn pattern.
128         (stack_protect_combined_test): New expander.
129         (stack_protect_combined_test_insn): New insn_and_split pattern.
130         (arm_stack_protect_test_insn): New insn pattern.
131         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
132         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
133         (UNSPEC_SP_TEST): Likewise.
134         * doc/md.texi (stack_protect_combined_set): Document new standard
135         pattern name.
136         (stack_protect_set): Clarify that the operand for guard's address is
137         legal.
138         (stack_protect_combined_test): Document new standard pattern name.
139         (stack_protect_test): Clarify that the operand for guard's address is
140         legal.
142 2018-11-22  Richard Biener  <rguenther@suse.de>
144         PR tree-optimization/88148
145         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
146         valueization callback parameter and handle it.
147         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
148         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
150 2018-11-22  Richard Biener  <rguenther@suse.de>
152         PR lto/87229
153         PR lto/88112
154         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
155         which can appear in size expressions.
156         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
157         CALL_EXPR_BY_DESCRIPTOR.
158         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
159         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
160         CALL_EXPR_BY_DESCRIPTOR.
161         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
163         Revert
164         PR lto/87229
165         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
166         sizepos values.
168 2018-11-22  Richard Biener  <rguenther@suse.de>
170         PR tree-optimization/88069
171         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
172         apply to default defs.
174 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
176         PR target/85644
177         PR target/86832
178         * config/i386/i386.c (ix86_option_override_internal): Default
179         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
180         is defined.
181         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
182         stack_protect_test, stack_protect_test_<mode>): Use empty condition
183         instead of TARGET_SSP_TLS_GUARD.
185 2018-11-22  Martin Liska  <mliska@suse.cz>
187         PR sanitizer/88017
188         * config/i386/i386.c (ix86_option_override_internal): Error
189         about usage -mabi=ms and -fsanitize=thread.
191 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
193         Revert the revert:
194         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
196         Revert:
197         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
199         * lra-spills.c (lra_final_code_change): Remove useless move insns.
201 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
203         PR middle-end/88129
204         * function.c (expand_function_end): Do not emit extra blockage insn.
206 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
208         PR target/85667
209         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
210         of FIRST_SSE_REG for 4 or 8 byte modes.
212 2018-11-21  Jan Hubicka  <jh@suse.cz>
214         PR lto/87957
215         * tree.c (fld_decl_context): Break out from ...
216         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
217         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
218         (fld_incomplete_type_of): Build copy of TYP_DECL.
219         * ipa-devirt.c (free_enum_values): Rename to ...
220         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
221         and TREE_TYPEs of TYPE_DECLs.
222         (get_odr_type): Initialize odr_vtable_hash if needed.
224 2018-11-21  Alexandre Oliva <oliva@adacore.com>
226         * final.c (compute_discriminator): Declare.  Renamed from...
227         (maybe_set_discriminator): ... this.  Set and return a local.
228         (override_discriminator): New.
229         (final_scan_insn_1): Set it.
230         (notice_source_line): Adjust.  Always set discriminator.
232 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
234         PR target/87839
235         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
236         rIJ constraint for aarch64_plus_operand rather than rn.
238 2018-11-21  Renlin Li  <renlin.li@arm.com>
240         PR middle-end/84877
241         * explow.h (get_dynamic_stack_size): Declare it as external.
242         * explow.c (record_new_stack_level): Remove function static attribute.
243         * function.c (assign_stack_local_1): Dynamically align the stack slot
244         addr for parameter copy on the stack.
246 2018-11-21  Richard Biener  <rguenther@suse.de>
248         PR bootstrap/88133
249         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
250         * Makefile.in (bitmap.o-warn): Remove again.
252 2018-11-20  Jeff Law  <law@redhat.com>
254         PR tree-optimization/88069
255         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
256         degenerate virtual PHIs.
258 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
260         PR target/87317
261         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
262         nonimmediate_operand with register_operand.
263         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
264         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
265         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
266         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
267         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
268         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
269         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
270         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
271         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
272         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
273         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
274         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
275         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
276         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
277         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
278         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
279         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
280         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
281         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
282         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
283         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
284         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
285         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
286         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
287         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
288         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
289         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
291 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
293         * read-rtl.c (apply_subst_iterator): Handle define_split and
294         define_insn_and_split.
296 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
298         PR rtl-optimization/87817
299         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
300         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
301         in the pattern to avoid triggering UB when operands[2] is zero.
302         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
303         (*tbm_bextri_<mode>): ... this.
305 2018-11-21  Tom de Vries  <tdevries@suse.de>
307         PR driver/79855
308         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
309         period.
311 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
313         * configure.ac: Add check for Binutils to determine whether vector
314         load/store alignments hints are being supported.
315         * config.in: Regenerate.
316         * configure: Regenerate.
317         * config/s390/s390.c (print_operand): Support new output
318         modifier A.
319         * config/s390/s390.md ("movti"): Append alignment hint output
320         using the new output modifier 'A'.
321         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
322         ("*vec_ti_to_v1ti"): Likewise.
324 2018-11-20  Martin Sebor  <msebor@redhat.com>
326         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
328 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
330         PR lto/84044
331         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
332         compare ENUM values.
334 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
336         PR tree-optimization/87895
337         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
338         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
339         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
340         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
341         has any preds and always use EDGE_PRED (, 0).
343 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
345         PR target/88070
346         * mode-switching.c (create_pre_exit): After reload, always split the
347         fallthrough edge to the exit block.
349 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
351         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
352         * tree.c (build_array_type_1): Forward declare.
353         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
354         (fld_type_variant): Likewise.
355         (fld_simplified_types): New hash.
356         (fld_process_array_type): New function.
357         (fld_incomplete_type_of): Handle array and enumeration types.
358         (fld_simplified_type): Handle simplification of arrays.
359         (free_lang_data): Allocate and free simplified types hash.
361 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
363         PR lto/87957
364         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
365         when sybtype already violates ODR.
366         (get_odr_type): Do not ICE when insert is false and type duplicate
367         is not registered yet.
368         (register_odr_type): Be sure to register subtypes first.
370 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
372         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
373         DImode parts of the target operand.
375 2018-11-20  Nathan Sidwell  <nathan@acm.org>
377         PR 87926
378         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
380 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
382         PR lto/87997
383         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
384         type checking to...
385         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
387 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
389         PR ipa/87706
390         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
391         * ipa.c (possible_inline_candidate_p): Break out from ..
392         (process_references): ... here ; drop before_inlining_p;
393         cleanup handling of alises.
394         (walk_polymorphic_call_targets): Likewise.
395         (symbol_table::remove_unreachable_nodes): Likewise.
396         * passes.c (pass_data_ipa_remove_symbols): New structure.
397         (pass_ipa_remove_symbols): New pass.
398         (make_pass_ipa_remove_symbols): New function.
399         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
400         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
402 2018-11-20  Richard Biener  <rguenther@suse.de>
404         * tree-vect-stmts.c (vectorizable_condition): Do not get
405         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
406         pointless vect_is_simple_use calls.
408 2018-11-20  Richard Biener  <rguenther@suse.de>
410         PR tree-optimization/88074
411         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
412         backedges.
414 2018-11-20  Richard Biener  <rguenther@suse.de>
416         PR tree-optimization/88069
417         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
418         virtual PHI arguments.
420 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
422         PR target/88083
423         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
424         involved.
425         * rtl.h (contains_constant_pool_address_p): New function.
426         * rtlanal.c (contains_constant_pool_address_p): Likewise.
428 2018-11-20  Richard Biener  <rguenther@suse.de>
430         PR middle-end/83215
431         * alias.c (component_uses_parent_alias_set_from): Remove
432         alias-set zero and TYPE_TYPELESS_STORAGE case both already
433         handled in other ways.
435 2018-11-20  Richard Biener  <rguenther@suse.de>
437         PR tree-optimization/88087
438         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
439         call fntype.
440         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
441         call fntype.
443 2018-11-20  Richard Biener  <rguenther@suse.de>
445         PR middle-end/88089
446         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
448 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
450         PR lto/87957
451         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
453 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
455         PR driver/50250
456         * doc/invoke.texi (Link Options): Mention shared libraries
457         in documentation for the -l option.  Simplify discussion and
458         point to the system linker documentation for details.
460 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
462         PR rtl-optimization/88033
463         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
464         to itself.  Use HARD_REGISTER_NUM_P.
466 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
468         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
469         of the loop if vf is non-constant.
471 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
473         PR tree-optimization/87025
474         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
475         immediately after creating the scope optinfo.
476         (selftest::test_pr87025): New function.
477         (selftest::dumpfile_c_tests): Call it.
478         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
479         that we're not popping the top-level records array.
480         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
481         not changing the kind of a "scope" optinfo.
483 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
485         PR tree-optimization/87025
486         * dump-context.h: Include "optinfo.h".
487         (class optrecord_json_writer): New forward decl.
488         (dump_context::forcibly_enable_optinfo_p): Delete.
489         (dump_context::optinfo_enabled_p): New member function.
490         (dump_context::optimization_records_enabled_p): New member
491         function.
492         (dump_context::set_json_writer): New member function.
493         (dump_context::emit_optinfo): New member function.
494         (dump_context::m_forcibly_enable_optinfo): Delete.
495         (dump_context::m_json_writer): New member data.
496         * dumpfile.c (dump_context::set_json_writer): New member function.
497         (dump_context::finish_any_json_writer): New member function.
498         (dump_context::end_scope): Replace call to
499         optimization_records_maybe_pop_dump_scope with call to
500         m_json_writer->pop_scope.
501         (dump_context::optinfo_enabled_p): New member function.
502         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
503         to dump_context::emit_optinfo.
504         (dump_context::emit_optinfo): New member function.
505         (temp_dump_context::temp_dump_context): Replace
506         m_forcibly_enable_optinfo with call to set_json_writer.
507         (temp_dump_context::~temp_dump_context): Clean up any json writer.
508         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
509         optinfo-emit-json.h
510         (the_json_writer): Delete.
511         (optimization_records_start): Delete.
512         (optimization_records_finish): Delete.
513         (optimization_records_enabled_p): Delete, in favor of
514         dump_context::optimization_records_enabled_p.
515         (optimization_records_maybe_record_optinfo): Delete.
516         (optimization_records_maybe_pop_dump_scope): Delete.
517         * optinfo-emit-json.h: Include "json.h".  Delete forward
518         decl of opt_pass.
519         (optimization_records_start): Delete.
520         (optimization_records_finish): Delete.
521         (optimization_records_enabled_p): Delete.
522         (optimization_records_maybe_record_optinfo): Delete.
523         (optimization_records_maybe_pop_dump_scope): Delete.
524         (class optrecord_json_writer): Move here from
525         optinfo-emit-json.cc.
526         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
527         to optinfo::emit with call to dump_context::emit_optinfo.
528         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
529         (optinfo_enabled_p): Delete, in favor of
530         dump_context::optinfo_enabled_p.
531         (optinfo_wants_inlining_info_p): Update for conversion o
532         optimization_records_enabled_p to a member function of
533         dump_context.
534         * optinfo.h (optinfo_enabled_p): Delete, in favor of
535         dump_context::optinfo_enabled_p.
536         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
537         * toplev.c: Include "dump-context.h".
538         (compile_file): Replace call to optimization_records_finish with
539         dump_context::finish_any_json_writer.
540         (do_compile): Replace call to optimization_records_start with
541         conditionally creating a optrecord_json_writer for the
542         dump_context.
544 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
546         PR tree-optimization/88071
547         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
548         of false as last argument to gsi_remove.
549         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
550         false as last argument to gsi_replace.
552         PR debug/87039
553         * omp-expand.c: Don't include debug.h.
554         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
555         from innermost outer parallel, task, teams or target that has a
556         child_fn set, or, if there is no such outer region, use
557         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
558         whether a suitable BLOCK is found or not.
559         (expand_parallel_call, expand_teams_call): Don't call
560         adjust_context_and_scope here.
561         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
562         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
563         adjust_context_and_scope.
564         * dwarf2out.c (dwarf2out_early_global_decl): For
565         decl_function_context recurse instead of calling dwarf2out_decl.
567 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
569         PR rtl-optimization/85925
570         * rtl.h (word_register_operation_p): New predicate.
571         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
572         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
573         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
574         (num_sign_bit_copies1): Likewise.
576 2018-11-19  Richard Biener  <rguenther@suse.de>
578         PR lto/87229
579         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
580         sizepos values.
582 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
584         * compare-elim.c (struct comparison): Add not_in_a field.
585         (is_not): New static function.
586         (strip_not): Likewise.
587         (conforming_compare): Handle a NOT in the first operand.
588         (can_eliminate_compare): Likewise.
589         (find_comparison_dom_walker::before_dom_children): Likewise.
590         (try_eliminate_compare): Likewise.
591         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
592         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
594 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
596         * doc/extend.texi (Common Type Attributes): Fix typos.
598 2018-11-19  Martin Liska  <mliska@suse.cz>
600         PR gcov-profile/88045
601         * coverage.c (coverage_begin_function): Add assert.
603 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
605         PR other/40498
606         * doc/extend.texi (Common Function Attributes): Document that
607         no_instrument_function applies to -p and -pg, too.
608         * doc/invoke.texi (Instrumentation Options): Add cross-references
609         to docs for -p, -pg, and -finstrument-functions.
611 2018-11-18  Alan Modra  <amodra@gmail.com>
613         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
614         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
615         (GNU_USER_TARGET_CRTN): Define.
616         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
617         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
619 2018-11-18  Alan Modra  <amodra@gmail.com>
621         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
622         !HAVE_LD_PIE variant.
623         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
625 2018-11-17  Nick Clifton  <nickc@redhat.com>
626             Sandra Loosemore  <sandra@codesourcery.com>
628         PR driver/31357
629         * doc/invoke.texi (Overall Options): Document that --help and
630         --help=<value> options cannot be combined.
632 2018-11-17  Jim Wilson  <jimw@sifive.com>
634         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
635         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
636         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
637         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
638         to step2 if frame_pointer_needed and step1 is 0.
640 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
642         PR c++/4225
643         * doc/cpp.texi (System Headers): Add note about implicit
644         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
646 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
648         PR c++/4025
649         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
650         -fno-implicit-templates.
652 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
654         PR tree-optimization/87546
655         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
656         min_precision variable, initially set it to orig_precision, only does
657         something if op_type's precision is <= min_precision and update
658         min_precision whenever calling set_op.
660 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
662         PR ipa/87957
663         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
665 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
667         * doc/invoke.texi (Option Summary): Fix whitespace and line
668         breaks in @gccoptlist environments.
669         (Warning Options): Likewise.
670         (Optimize Options): Likewise.
671         (PowerPC SPE Options): Likewise.
672         (RS/6000 and PowerPC Options): Likewise.
674 2018-11-16  Jeff Law  <law@redhat.com>
676         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
677         gcc_assert.
679 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
681         PR middle-end/23197
682         * doc/invoke.texi (Optimize Options): Update options enabled by
683         fprofile-generate, -fprofile-use, and -fauto-profile.
685 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
687         PR target/87927
688         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
689         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
690         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
691         the byte_op hook.
692         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
693         * targhooks.c (default_print_patchable_function_entry): Assert
694         asm_int_op does not return a NULL string.
695         * varasm.c (integer_asm_op): Return the op for a partial int type
696         when the requested size does not correspond to an integer type.
697         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
698         * doc/tm.texi: Regenerate.
700 2018-11-12  Jason Merrill  <jason@redhat.com>
702         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
704 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
706         * config/rs6000/constraints.md (wF constraint): Remove power9
707         fusion documentation.  Just document wF's use for power8 fusion.
708         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
709         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
710         fusion support to require the upper 12 bits to be all 0's or all
711         1's.
712         (fusion_gpr_mem_load): Add comment.
713         (fusion_addis_mem_combo_load): Remove power9 fusion support.
714         (fusion_addis_mem_combo_store): Delete.
715         (fusion_offsettable_mem_operand): Delete.
716         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
717         power8 fusion here.
718         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
719         (POWERPC_MASKS): Delete power9 fusion.
720         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
721         (fusion_p9_p): Delete.
722         (expand_fusion_p9_load): Delete.
723         (expand_fusion_p9_store): Delete.
724         (emit_fusion_p9_load): Delete.
725         (emit_fusion_p9_store): Delete.
726         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
727         fusion support.
728         (rs6000_option_override_internal): Set power8 fusion based on
729         whether we are tuning for power8.  Delete power9 fusion support.
730         (rs6000_opt_masks): Delete -mpower9-fusion switch.
731         (emit_fusion_load): Rename emit_fusion_load_store to
732         emit_fusion_load, and drop fusion store support.  Update callers.
733         (emit_fusion_load_store): Likewise.
734         (emit_fusion_gpr_load): Likewise.
735         (fusion_p9_p): Delete.
736         (expand_fusion_p9_load): Delete.
737         (expand_fusion_p9_store): Delete.
738         (emit_fusion_p9_load): Delete.
739         (emit_fusion_p9_store): Delete.
740         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
741         (GPR_FUSION): Delete.
742         (FPR_FUSION): Delete.
743         (power9 fusion peephole2s): Delete.
744         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
745         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
746         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
747         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
748         (fusion_p9_<mode>_constant): Delete.
749         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
750         power9 fusion switch.
751         * doc/md.texi (PowerPC constraints): Update wF constraint
752         documentation for power8 fusion only.
754 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
756         PR rtl-optimization/87475
757         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
758         CROSSING_JUMP_P insns.
759         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
761 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
763         PR target/88051
764         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
765         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
766         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
768 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
770         PR middle-end/88032
771         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
772         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
774 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
776         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
777         location_t parameter.
778         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
779         (switch_decision_tree::do_jump_if_equal): Likewise.
780         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
781         of switch statement to emit_case_nodes.
782         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
783         set it on the newly built GIMPLE comparison statement.
784         (switch_decision_tree::do_jump_if_equal): Likewise.
785         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
786         in calls to do_jump_if_equal as well as recursive calls.
788 2018-11-16  Andi Kleen  <ak@linux.intel.com>
790         * config/i386/i386.md: Allow memory operands to ptwrite.
792 2018-11-16  Michael Matz  <matz@suse.de>
794         * system.h (PRsa): New macro.
795         (SIZE_AMOUNT): Cast number to uint64_t.
796         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
797         (pool_usage::dump_footer): Likewise and also use PRIu64.
798         * bitmap.h (bitmap_usage::dump): Likewise.
799         * ggc-common.c (ggc_usage::dump): Likewise.
800         * ggc-page.c (ggc_print_statistics): Likewise.
801         * input.c (dump_line_table_statistics): Likewise.
802         * mem-stats.h (mem_usage::dump): Likewise.
803         (mem_usage::dump_footer): Likewise.
804         * rtl.c (dump_rtx_statistics): Likewise.
805         * tree-cfg.c (dump_cfg_stats): Likewise.
806         * tree-dfa.c (dump_dfa_stats): Likewise.
807         * tree-phinodes.c (phinodes_print_statistics): Likewise.
808         * tree-ssanames (ssanames_print_statistics): Likewise.
809         * vec.c (vec_usage::dump): Likewise.
810         (vec_usage::dump_footer): Likewise.
812 2018-11-16  Richard Biener  <rguenther@suse.de>
814         PR tree-optimization/88011
815         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
816         replacing set_value_range_to_undefined and
817         set_value_range_to_varying with method calls.
819 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
821         * config/s390/s390.md
822         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
823         assembler.
824         (*r<noxa>sbg_<mode>_sll): Likewise.
825         (*r<noxa>sbg_<mode>_srl): Likewise.
826         (*r<noxa>sbg_sidi_srl): New pattern.
828 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
830         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
831         Default to TARGET_VXWORKS_RTP.
832         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
833         of TARGET_VXWORKS_RTP.
834         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
835         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
836         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
838 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
840         PR c++/25759
841         * doc/extend.texi (Common Type Attributes): Make it explicit
842         that attribute "packed" can apply to C++ classes.
844 2018-11-15  Martin Sebor  <msebor@redhat.com>
846         PR c++/87541
847         PR c++/87542
848         * tree.c (type_argument_type): New function.
849         * tree.h (type_argument_type): Declare it.
850         * gcc/doc/extend.texi (alloc_align): Update and clarify.
851         (alloc_size, nonnull, sentinel): Same.
853 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
854             Kwok Cheung Yeung  <kcy@codesourcery.com>
856         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
857         int_mode_for_size fails.
858         (vectorizable_load): Likewise.
860 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
862         * doc/ux.texi (Group logically-related diagnostics): Move
863         discussion of auto_diagnostic_group into this new subsection.
864         Give an example of where this grouping is used.
866 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
868         PR other/19165
869         * Makefile.in (OBJS): Move json.o to...
870         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
871         * common.opt (fdiagnostics-format=): New option.
872         (diagnostics_output_format): New enum.
873         * diagnostic-format-json.cc: New file.
874         * diagnostic.c (default_diagnostic_final_cb): New function, taken
875         from start of diagnostic_finish.
876         (diagnostic_initialize): Initialize final_cb to
877         default_diagnostic_final_cb.
878         (diagnostic_finish): Move "being treated as errors" messages to
879         default_diagnostic_final_cb.  Call any final_cb.
880         (default_diagnostic_finalizer): Add diagnostic_t param.
881         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
882         diagnostic_finalizer callback.
883         * diagnostic.h (enum diagnostics_output_format): New enum.
884         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
885         (struct diagnostic_context): Add "final_cb".
886         (default_diagnostic_finalizer): Add diagnostic_t param.
887         (diagnostic_output_format_init): New decl.
888         * doc/invoke.texi (-fdiagnostics-format): New option.
889         * dwarf2out.c (gen_producer_string): Ignore
890         OPT_fdiagnostics_format_.
891         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
892         * lto-wrapper.c (append_diag_options): Ignore it.
893         * opts.c (common_handle_option): Handle it.
895 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
897         PR tree-optimization/88015
898         * graphite-isl-ast-to-gimple.c
899         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
900         for dump_enabled_p.
901         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
903 2018-11-15  Richard Biener  <rguenther@suse.de>
905         PR middle-end/88029
906         * gimple.c (gimple_call_flags): Union flags from decl, type
907         and call fntype.
908         * trans-mem.c (is_tm_pure_call): Simplify.
910 2018-11-15  Richard Biener  <rguenther@suse.de>
912         PR tree-optimization/88031
913         * tree-vect-loop.c (vectorizable_reduction): Move check
914         for multiple types earlier so we get the expected dump.
915         Simplify calls to vectorizable_condition.
916         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
917         * tree-vect-stmts.c (vectorizable_condition): Instead of
918         reduc_def and reduc_index take just a flag.  Simplify
919         code-generation now that we can rely on the defs being set up.
920         (vectorizable_comparison): Remove unused argument.
922 2018-11-15  Richard Biener  <rguenther@suse.de>
924         PR tree-optimization/88030
925         * tree-complex.c (need_eh_cleanup): New global.
926         (update_complex_assignment): Mark blocks that need EH update.
927         (expand_complex_comparison): Likewise.
928         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
929         perform EH cleanup and schedule CFG cleanup if that did anything.
931 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
933         PR rtl-optimization/88018
934         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
935         on the presence of fallthru edge, rather than if it is a USE or not.
937 2018-11-15  Richard Biener  <rguenther@suse.de>
939         PR middle-end/87917
940         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
941         analyze_subscript_affine_affine properly.
943 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
945         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
946         (GLIBC_DYNAMIC_LINKER): Define.
947         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
949 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
951         PR tree-optimization/84648
952         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
953         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
954         by calling adjust_cond_for_loop_until_wrap.
956 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
958         PR other/56334
959         * doc/extend.texi (Common Function Attributes): Clarify linker
960         restrictions on "aligned" attribute.
961         (Common Variable Attributes): Likewise.  Mention that linker
962         restrictions don't apply to stack-allocated variables.
964 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
966         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
967         as an lvalue.
969 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
971         PR lto/55102
972         PR lto/56700
973         * doc/invoke.texi (Optimize Options): Remove bad example about
974         interaction between -flto and -O.  Replace it with a note that
975         you need to compile with -O and not just link.  Copy-edit -flto
976         discussion to reduce verbiage and improve flow.
978 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
980         PR middle-end/59658
981         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
982         also suppress many optimizations.  Alphabetize option lists for
983         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
984         correct documentation for those options to say that.
985         * opts.c (default_options_table): Sort table by level and option
986         name, to make it easier to correlate to the manual.
988 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
990         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
991         printing of the names of segment registers.
993 2018-11-14  Nathan Sidwell  <nathan@acm.org>
995         PR debug/88006
996         PR debug/87462
997         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
998         type list.
1000 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
1002         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
1004 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1006         P1236R1 - Signed integers are two's complement
1007         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
1009         PR bootstrap/86739
1010         * hash-map.h (hash_map::iterator::reference_pair): New class.
1011         (hash_map::iterator::operator*): Return it rather than std::pair.
1013 2018-11-14  Jeff Law  <law@redhat.com>
1015         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
1016         iff the operand is a constant.
1018 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
1020         * gimple-ssa-evrp-analyze.c
1021         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
1022         ignore_equivs_equal_p to equal_p.
1023         * ipa-cp.c (meet_with_1): Use equal_p instead of
1024         ignore_equivs_equal_p.
1025         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
1026         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
1027         (value_range::operator==): Remove.
1028         (value_range::operator!=): Remove.
1029         (vrp_prop::visit_stmt): Use equal_p.
1030         * tree-vrp.h (value_range): Remove operator==, operator!=,
1031         ignore_equivs_equal_p.
1032         * vr-values.c (update_value_range): Use equal_p.
1034 2018-11-14  Michael Matz  <matz@suse.de>
1036         PR middle-end/86575
1037         * gimplify.c (collect_fallthrough_labels): Add new argument,
1038         return location via that, don't modify statements.
1039         (warn_implicit_fallthrough_r): Adjust call, don't use
1040         statement location directly.
1042 2018-11-14  Richard Biener  <rguenther@suse.de>
1044         PR middle-end/87985
1045         * tree-data-ref.c (split_constant_offset): Add wrapper
1046         allocating a cache hash-map.
1047         (split_constant_offset_1): Cache results of expanding
1048         expressions from SSA def stmts.
1050 2018-11-14  Richard Biener  <rguenther@suse.de>
1052         PR middle-end/88021
1053         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
1054         to lambda_int.
1055         (lambda_vector_mult_const): Likewise.
1056         (lambda_matrix_right_hermite): Use lambda_int temporaries.
1058 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
1059             Jackson Woodruff  <jackson.woodruff@arm.com>
1061         PR 71026/tree-optimization
1062         * match.pd: Simplify floating point comparisons.
1064 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1066         PR rtl-optimization/87817
1067         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
1068         last argument has low 8 bits clear, fold to 0.
1070 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
1072         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
1073         MAX_OFILE_ALIGNMENT as needed.
1075 2018-11-14  Richard Biener  <rguenther@suse.de>
1077         PR tree-optimization/88019
1078         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
1079         COND_EXPR arguments to non-trapping overflow.
1081 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
1083         * config/arc/arc.h (reg_class): Reorder registers classes, remove
1084         unused register classes.
1085         (REG_CLASS_NAMES): Likewise.
1086         (REG_CLASS_CONTENTS): Likewise.
1087         (FIXED_REGISTERS): Make lp_count fixed.
1088         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
1089         (PROGRAM_COUNTER_REGNO): Remove.
1090         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
1091         register classes, use constants for register numbers, remove
1092         reg_class_contents references.
1093         (arc_process_double_reg_moves): Add asserts.
1094         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
1095         lra_in_progress predicate.
1096         (arc_init_reg_tables): Remove unused register classes.
1097         (arc_register_move_cost): Likewise.
1098         (arc_preferred_reload_class): Likewise.
1099         (hwloop_optimize): Update rtx patterns involving lp_count register.
1100         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
1101         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
1102         CHEAP_CORE_REGS.  Former one will be used for LRA.
1103         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
1104         one will be used for LRA.
1105         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
1106         one will be used for LRA.
1107         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
1108         Former one will be used for LRA.
1109         ("f"): Delete constraint.
1110         ("k"): Likewise.
1111         ("e"): Likewise.
1112         ("l"): Change it from register constraint to constraint.
1113         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
1114         (movhi_insn): Likewise.
1115         (movsi_insn): Update pattern.
1116         (arc_lp): Likewise.
1117         (dbnz): Likewise.
1118         (stack_tie): Remove 'b' constraint letter.
1119         (R4_REG): Define.
1120         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
1121         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
1122         (R57_REG, R59_REG, PCL_REG): Likewise.
1123         (ILINK1_REGNUM): Renamed to ILINK1_REG.
1124         (ILINK2_REGNUM): Renamed to ILINK2_REG.
1125         (Rgp): Remove.
1126         (SP_REGS): Likewise.
1127         (Rcw): Remove unused reg classes.
1128         * config/arc/predicates.md (dest_reg_operand): Just default on
1129         register_operand predicate.
1130         (mpy_dest_reg_operand): Likewise.
1131         (move_dest_operand): Use macros instead of constants.
1132         * config/arc/arc.opt (mlra): Switch to lra as default.
1134 2018-11-14  Richard Biener  <rguenther@suse.de>
1136         PR tree-optimization/87974
1137         * tree-vect-loop.c (vectorizable_reduction): When computing
1138         the vectorized reduction PHI vector type ignore constant
1139         and external defs.
1141 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1143         PR tree-optimization/87977
1144         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
1145         stmt, build a new one and replace the old one with it.  Formatting fix.
1146         Call release_ssa_name (x) if !has_other_use and !delete_div.
1147         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
1148         verify lhs of stmt is still def.
1150 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
1152         PR rtl-optimization/87507
1153         * lower-subreg.c (operand_for_swap_move_operator): New function.
1154         (simple_move): Strip simple operators.
1155         (find_pseudo_copy): Likewise.
1156         (resolve_operand_for_swap_move_operator): New function.
1157         (resolve_simple_move): Strip simple operators and swap operands.
1159 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
1161         PR tree-optimization/87898
1162         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
1163         (ipa_simd_modify_function_body): Remove debug stmts where the first
1164         argument was changed into a non-decl.
1166 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
1168         PR rtl-optimization/87899
1169         * lra-lives.c (start_living): Update white space in comment.
1170         (enum point_type): New.
1171         (sparseset_contains_pseudos_p): New function.
1172         (update_pseudo_point): Likewise.
1173         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
1174         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
1175         handling.  Move early exit after adding conflicts.
1176         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
1177         if regno is already live.  Remove all handling of program points.
1178         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
1179         after adding conflicts.  Remove all handling of program points and
1180         ignore_reg_for_conflicts.
1181         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
1182         and do not guard call to mark_pseudo_live.
1183         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
1184         and do not guard call to mark_pseudo_dead.
1185         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
1186         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
1187         Use new function update_pseudo_point.  Handle register copies by
1188         removing the source register from the live set.  Handle INOUT operands.
1189         Update to the next program point using the unused_set, dead_set and
1190         start_dying sets.
1191         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
1193 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1195         * builtins.c: Replace "source_location" with "location_t".
1196         * diagnostic-show-locus.c: Likewise.
1197         * diagnostic.c: Likewise.
1198         * dumpfile.c: Likewise.
1199         * gcc-rich-location.h: Likewise.
1200         * genmatch.c: Likewise.
1201         * gimple.h: Likewise.
1202         * gimplify.c: Likewise.
1203         * input.c: Likewise.
1204         * input.h: Likewise.  Eliminate the typedef.
1205         * omp-expand.c: Likewise.
1206         * selftest.h: Likewise.
1207         * substring-locations.h (get_source_location_for_substring):
1208         Rename to...
1209         (get_location_within_string): ...this.
1210         * tree-cfg.c: Replace "source_location" with "location_t".
1211         * tree-cfgcleanup.c: Likewise.
1212         * tree-diagnostic.c: Likewise.
1213         * tree-into-ssa.c: Likewise.
1214         * tree-outof-ssa.c: Likewise.
1215         * tree-parloops.c: Likewise.
1216         * tree-phinodes.c: Likewise.
1217         * tree-phinodes.h: Likewise.
1218         * tree-ssa-loop-ivopts.c: Likewise.
1219         * tree-ssa-loop-manip.c: Likewise.
1220         * tree-ssa-phiopt.c: Likewise.
1221         * tree-ssa-phiprop.c: Likewise.
1222         * tree-ssa-threadupdate.c: Likewise.
1223         * tree-ssa.c: Likewise.
1224         * tree-ssa.h: Likewise.
1225         * tree-vect-loop-manip.c: Likewise.
1227 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1229         * doc/extend.texi: Fix typo in the weakref description.
1231 2018-11-13  Richard Biener  <rguenther@suse.de>
1233         PR tree-optimization/86991
1234         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
1235         group building until we have successfully detected the SLP
1236         reduction.
1237         (vect_is_simple_reduction): Remove fixup code here.
1239 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1241         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
1242         (dump_gimple_stmt): Use it.
1243         (dump_gimple_stmt_loc): Likewise.
1244         (dump_gimple_expr): Likewise.
1245         (dump_gimple_expr_loc): Likewise.
1246         (dump_generic_expr): Likewise.
1247         (dump_generic_expr_loc): Likewise.
1248         (dump_printf): Likewise.
1249         (dump_printf_loc): Likewise.
1250         (dump_dec): Likewise.
1251         (dump_dec): Likewise.
1252         (dump_hex): Likewise.
1253         (dump_symtab_node): Likewise.
1254         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
1255         Guard dump call with dump_enabled_p.
1256         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
1257         * graphite-optimize-isl.c (optimize_isl): Likewise.
1258         * graphite.c (graphite_transform_loops): Likewise.
1259         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
1260         * tree-parloops.c (parallelize_loops): Likewise.
1261         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
1262         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
1263         (vect_prune_runtime_alias_test_list): Likewise.
1264         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
1265         (vect_estimate_min_profitable_iters): Likewise.
1266         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
1267         (vect_build_slp_tree_2): Likewise.
1268         (vect_supported_load_permutation_p): Likewise.
1269         (vect_slp_analyze_operations): Likewise.
1270         (vect_slp_analyze_bb_1): Likewise.
1271         (vect_slp_bb): Likewise.
1272         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
1273         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
1274         (pass_slp_vectorize::execute): Likewise.
1275         (increase_alignment): Likewise.
1277 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1279         PR ipa/87955
1280         * ipa-inline.c (report_inline_failed_reason): Guard calls to
1281         cl_target_option_print_diff and cl_optimization_print_diff with
1282         if (dump_file).
1284 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1286         * doc/invoke.texi (-fsave-optimization-record): Note that the
1287         output is compressed.
1288         * optinfo-emit-json.cc: Include <zlib.h>.
1289         (optrecord_json_writer::write): Compress the output.
1291 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
1293         * tree-vrp.c (value_range_base::dump): Dump type.
1294         Do not use INF nomenclature for 1-bit types.
1295         (dump_value_range): Group all variants to common dumping code.
1296         (debug): New overloaded functions for value_ranges.
1297         (value_range_base::dump): Remove no argument version.
1298         (value_range::dump): Same.
1300 2018-11-13  Richard Biener  <rguenther@suse.de>
1302         PR tree-optimization/87931
1303         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
1304         nested cycles we support to latch computations vectorizable_reduction
1305         handles.
1307 2018-11-13  Martin Liska  <mliska@suse.cz>
1309         PR tree-optimization/87885
1310         * cfghooks.c (account_profile_record): Rename
1311         to ...
1312         (profile_record_check_consistency): ... this.
1313         Calculate missing num_mismatched_freq_in.
1314         (profile_record_account_profile): New function
1315         that calculates time and size of a function.
1316         * cfghooks.h (struct profile_record): Remove
1317         all tuples.
1318         (struct cfg_hooks): Remove after_pass flag.
1319         (account_profile_record): Rename to ...
1320         (profile_record_check_consistency): ... this.
1321         (profile_record_account_profile): New.
1322         * cfgrtl.c (rtl_account_profile_record): Remove
1323         after_pass flag.
1324         * passes.c (check_profile_consistency): Do only
1325         checking.
1326         (account_profile): Calculate size and time of
1327         function only.
1328         (pass_manager::dump_profile_report): Reformat
1329         output.
1330         (execute_one_ipa_transform_pass): Call
1331         consistency check before clean upand call account_profile
1332         after a clean up is done.
1333         (execute_one_pass): Call check_profile_consistency and
1334         account_profile instead of using after_pass flag..
1335         * tree-cfg.c (gimple_account_profile_record): Likewise.
1337 2018-11-13  Martin Liska  <mliska@suse.cz>
1339         PR sanitizer/87930
1340         * config/i386/i386.c (ix86_option_override_internal): Error
1341         about usage -mabi=ms and -fsanitize={,kernel-}address.
1343 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1345         * config/aarch64/aarch64.c
1346         (aarch64_vectorize_preferred_vector_alignment): Change return type to
1347         poly_uint64.
1348         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
1349         alignment being a poly int.
1350         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
1351         return type to poly_uint64.
1352         * target.def (default_preferred_vector_alignment): Likewise.
1353         * targhooks.c (default_preferred_vector_alignment): Likewise.
1354         * targhooks.h (default_preferred_vector_alignment): Likewise.
1355         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
1356         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
1357         poly int.
1358         (vect_update_misalignment_for_peel): Likewise.
1359         (vect_enhance_data_refs_alignment): Likewise.
1360         (vect_find_same_alignment_drs): Likewise.
1361         (vect_duplicate_ssa_name_ptr_info): Likewise.
1362         (vect_setup_realignment): Likewise.
1363         (vect_can_force_dr_alignment_p): Change alignment parameter type to
1364         poly_uint64.
1365         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
1366         mask with a compile time variable vector alignment.
1367         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
1368         int.
1369         (vect_do_peeling): Exit early if vector alignment is not constant.
1370         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
1371         a poly int.
1372         (vectorizable_store): Likewise.
1373         (vectorizable_load): Likweise.
1374         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
1375         poly_uint64.
1376         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
1377         poly int.
1378         (vect_can_force_dr_alignment_p): Change alignment parameter type to
1379         poly_uint64.
1381 2018-11-13  Richard Biener  <rguenther@suse.de>
1383         PR tree-optimization/87962
1384         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
1385         detect outer reduction for disqualifying in-loop uses.
1387 2018-11-13  Richard Biener  <rguenther@suse.de>
1389         PR tree-optimization/87967
1390         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
1391         for constants for the scalar loop.
1393 2018-11-13  Alan Modra  <amodra@gmail.com>
1395         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
1396         integer constants when -mcmodel=medium.
1398 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1400         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
1401         * config/arc/elf.h (PROFILE_HOOK): Define.
1402         * config/arc/linux.h (PROFILE_HOOK): Likewise.
1404 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1406         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
1407         jump table data in the text section.
1409 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1411         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
1412         to fit the eh_return pattern.
1413         * config/arc/arc.md (eh_return): Define.
1414         (VUNSPEC_ARC_EH_RETURN): Likewise.
1415         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
1416         new implementation.
1417         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
1420 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1422         * common/config/arc/arc-common.c (arc_option_optimization_table):
1423         Millicode optimization is default on for size optimizations.
1424         * config/arc/arc-protos.h (arc_check_multi): New function.
1425         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
1426         (ENTER_LEAVE_START_REG): Define.
1427         (ENTER_LEAVE_END_REG): Likewise.
1428         (arc_override_options): Disable millicode when long calls option
1429         is on.
1430         (arc_frame_info): Change it from int to bool.
1431         (arc_compute_frame_size): Clean up.
1432         (arc_save_restore): Remove.
1433         (frame_save_reg): New function.
1434         (frame_restore_reg): Likewise.
1435         (arc_enter_leave_p): Likewise.
1436         (arc_save_callee_saves): Likewise.
1437         (arc_restore_callee_saves): Likewise.
1438         (arc_save_callee_enter): Likewise.
1439         (arc_restore_callee_leave): Likewise.
1440         (arc_save_callee_milli): Likewise.
1441         (arc_restore_callee_milli): Likewise.
1442         (arc_expand_prologue): Reimplement to emit enter/leave
1443         instructions.
1444         (arc_expand_epilogue): Likewise.
1445         (arc_check_multi): New function.
1446         * config/arc/arc.md (push_multi_fp): New pattern.
1447         (push_multi_fp_blink): Likewise.
1448         (pop_multi_fp): Likewise.
1449         (pop_multi_fp_blink): Likewise.
1450         (pop_multi_fp_ret): Likewise.
1451         (pop_multi_fp_blink_ret): Likewise.
1452         * config/arc/arc.opt (mmillicode): Update option.
1453         (mcode-density-frame): New option.
1454         * config/arc/predicates.md (push_multi_operand): New predicate.
1455         (pop_multi_operand): Likewise.
1456         * doc/invoke.texi (ARC): Update ARC options information.
1459 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1461         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
1462         * config/arc/arc.c (operands_ok_ldd_std): New function.
1463         (mem_ok_for_ldd_std): Likewise.
1464         (gen_operands_ldd_std): Likewise.
1465         * config/arc/arc.md: Add peephole2 rules for std/ldd.
1467 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
1469         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
1470         test on stack_usage_file.
1471         (lang_dependent_init): Do not open the .su file if generating LTO.
1473 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
1475         PR rtl-optimization/87918
1476         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
1477         simplify_gen_relational rather than simplify_gen_binary.
1479 2018-11-13  Richard Biener  <rguenther@suse.de>
1481         * tree-ssanames.h (set_range_info): Use value_range_base.
1482         (get_range_info): Likewise.
1483         * tree-ssanames.c (set_range_info): Likewise.
1484         (get_range_info): Likewise.
1485         * tree-vrp.c (value_range_base::union_helper): Split
1486         out common parts of value_range[_base]::union_.
1487         (value_range_base::union_): Update.
1488         (value_range::union_): Likewise.
1489         (determine_value_range_1): Use value_range_base.
1490         (determine_value_range): Likewise.
1491         * tree-vrp.h (value_range_base::union_helper): Move ...
1492         (value_range::union_helper): ... from here.
1494 2018-11-13  Alan Modra  <amodra@gmail.com>
1496         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
1497         offset for PRE_DEC.
1498         (rs6000_secondary_reload_gpr): Don't call find_replacement.
1500 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
1502         PR middle-end/59634
1503         * doc/invoke.texi (Optimize Options): Clarify that the
1504         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
1505         apply to data cache size.
1507 2018-11-13  Alan Modra  <amodra@gmail.com>
1509         * config/rs6000/predicates.md (logical_const_operand),
1510         (logical_operand): Correct comment.
1511         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
1513 2018-11-13  Alan Modra  <amodra@gmail.com>
1515         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
1517 2018-11-13  Alan Modra  <amodra@gmail.com>
1519         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
1520         description.
1521         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
1522         covered by alternative.
1523         (movcc_internal1): Ignore h for register preference.
1524         (mov<mode>_hardfloat64): Likewise.
1525         (mov<mode>_softfloat): Ignore c, l, h for register preference.
1527 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
1529         PR preprocessor/47823
1530         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
1531         documentation to...
1532         (Pragmas): ...here.
1533         * doc/extend.texi (Pragmas): Note additional pragmas documented
1534         in the CPP manual.
1536 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
1538         PR middle-end/86677
1539         PR middle-end/87528
1540         * tree-scalar-evolution.c (expression_expensive_p): Make
1541         BUILTIN POPCOUNT as expensive when backend does not define it.
1543 2018-11-12  Fredrik Noring  <noring@nocrew.org>
1545         * config.gcc: Update with-llsc defaults for MIPS r5900.
1547 2018-11-12  Martin Liska  <mliska@suse.cz>
1549         PR gcov-profile/87442
1550         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
1551         options.
1552         * doc/invoke.texi: Document them.
1553         * tree-profile.c (parse_profile_filter): New.
1554         (parse_profile_file_filtering): Likewise.
1555         (release_profile_file_filtering): Likewise.
1556         (include_source_file_for_profile): Likewise.
1557         (tree_profiling): Filter source files based on the
1558         newly added options.
1560 2018-11-12  Martin Liska  <mliska@suse.cz>
1562         PR target/87903
1563         * doc/extend.texi: Add missing values for __builtin_cpu_is and
1564         __builtin_cpu_supports for x86 target.
1566 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
1568         PR middle-end/21110
1569         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
1570         have to be Pmode.
1571         (Arithmetic): Likewise for "lo_sum".
1573 2018-11-12  Renlin Li  <renlin.li@arm.com>
1575         PR target/87815
1576         * dse.c (get_stored_val): Add check for compile-time constantness
1577         of gap.
1579 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
1581         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
1582         (ARMv8_5a): New fgroup.
1583         (armv8.5-a): New arch.
1584         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
1585         sb and predres.
1586         * config/arm/arm-tables.opt: Regenerate.
1587         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
1588         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
1589         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
1590         Add matching rules for -march=armv8.5-a and extensions.
1591         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
1592         Add sb and predres to all armv8-a except armv8.5-a.
1594 2018-11-12  Richard Biener  <rguenther@suse.de>
1596         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
1597         overload for single value.
1598         (value_range[_base]::set_nonnull): New.
1599         (value_range[_base]::set_null): Likewise.
1600         (value_range): Document bitmap copying behavior, mark
1601         copy constructor and assignment operator deleted.
1602         (value_range::move): New.
1603         (value_range::set_and_canonicalize): Default bitmap to zero.
1604         (set_value_range_to_nonnull): Remove.
1605         (set_value_range_to_null): Likewise.
1606         (set_value_range): Likewise.
1607         (set_value_range_to_value): Likewise.
1608         (extract_range_from_unary_expr): Work on value_range_base.
1609         (extract_range_from_binary_expr_1): Likewise.  Rename to...
1610         (extract_range_from_binary_expr): ... this.
1611         * tree-vrp.c (value_range::update): Clear equiv bitmap
1612         if required.
1613         (value_range::move): New, move equiv bitmap.
1614         (value_range_base::set_undefined): Avoid assignment.
1615         (value_range::set_undefined): Likewise.
1616         (value_range_base::set_varying): Likewise.
1617         (value_range::set_varying): Likewise.
1618         (set_value_range): Remove.
1619         (value_range_base::set): New overload for value.
1620         (value_range::set): Likewise.
1621         (set_value_range_to_nonnull): Remove.
1622         (value_range_base::set_nonnull): New.
1623         (value_range::set_nonnull): Likewise.
1624         (set_value_range_to_null): Remove.
1625         (value_range_base::set_null): New.
1626         (value_range::set_null): Likewise.
1627         (range_is_null): Work on value_range_base.
1628         (range_is_nonnull): Likewise.
1629         (ranges_from_anti_range): Likewise.
1630         (extract_range_into_wide_ints): Likewise.
1631         (extract_range_from_multiplicative_op): Likewise.
1632         (extract_range_from_binary_expr): Likewise.  Update for API changes.
1633         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
1634         handling.
1635         (value_range::intersect_helper): Avoid copy and assignment.
1636         (value_range::union_helper): Likewise.
1637         (determine_value_range_1): Adjust.
1638         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
1639         Avoid assignment by using move.
1640         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
1641         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1642         Likewise.
1643         * tree-ssanames.c (get_range_info): Likewise.
1644         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
1645         * vr-values.c (vr_values::get_value_range): Adjust.
1646         (vr_values::update_value_range): Likewise.
1647         (symbolic_range_based_on_p): Work on value_range_base.
1648         (vr_values::extract_range_from_binary_expr): Use value_range_base.
1649         (vr_values::extract_range_from_unary_expr): Likewise.
1650         (vr_values::extract_range_from_cond_expr): Avoid assignment.
1651         (vr_values::extract_range_from_comparison): Adjust.
1652         (vr_values::check_for_binary_op_overflow): Use value_range_base.
1653         (vr_values::extract_range_basic): Adjust.
1654         (vr_values::adjust_range_with_scev): Likewise.
1655         (vr_values::vrp_visit_assignment_or_call): Likewise.
1656         (vr_values::get_vr_for_comparison): Change API to avoid
1657         assignment and copy construction.
1658         (vr_values::compare_name_with_value): Adjust accordingly.
1659         (vr_values::compare_names): Likewise.
1660         (vr_values::extract_range_from_phi_node): Avoid assignment and
1661         bogus in-place modify of equiv bitmap.
1662         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
1663         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
1664         for extract_range_from_unary_expr API change.
1665         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
1667 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
1669         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
1671 2018-11-12  Richard Biener  <rguenther@suse.de>
1673         * tree-vrp.h (value_range_base::symbolic_p,
1674         value_range_base::constant_p, value_range_base::zero_p,
1675         value_range_base::singleton_p): Move from value_range.
1676         (value_range::dump): Add.
1677         * gimple-ssa-evrp-analyze.c
1678         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
1679         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
1680         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1681         Use set_varying.
1682         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
1683         (value_range::constant_p): Likewise.
1684         (value_range::singleton_p): Likewise.
1685         (value_range_base::dump): Add.
1686         (set_value_range_to_undefined): Remove.
1687         (set_value_range_to_varying): Likewise.
1688         (range_int_cst_p): Take value_range_base argument.
1689         (range_int_cst_singleton_p): Likewise.
1690         (value_range_constant_singleton): Likewise.
1691         (vrp_set_zero_nonzero_bits): Likewise.
1692         (extract_range_from_multiplicative_op): Use set_varying.
1693         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1694         (extract_range_from_unary_expr): Likewise.
1695         (dump_value_range_base): Change to overload of dump_value_range.
1696         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1697         (vrp_prop::visit_stmt): Likewise.
1698         (value_range::intersect_helper): Likewise.
1699         (value_range::union_helper): Likewise.
1700         (determine_value_range_1): Likewise.
1702 2018-11-12  Richard Biener  <rguenther@suse.de>
1704         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1705         (set_value_range_to_null): Likewise.
1706         * vr-values.c (vr_values::extract_range_from_comparison):
1707         Clear equiv for constant singleton ranges.
1709 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1711         * config/i386/sse.md: Combine VFIXUPIMM* patterns
1712         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1713         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1714         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1715         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1716         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1717         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1719 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1721         PR c/69502
1722         * doc/extend.texi (Common Type Attributes): For the align type
1723         attribute, copy language about decreasing alignment from the
1724         corresponding variable attribute.
1726 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1728         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1729         -frounding-math.
1731 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1733         PR c++/43105
1734         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1735         -frtti and -fno-rtti code.
1737 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1739         PR c/26366
1740         * doc/extend.texi (Other Builtins): Document probability associated
1741         with __builtin_expect.
1743 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
1745         PR target/87928
1746         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1747         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1748         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1749         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1751 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1753         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1755 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1757         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1759 2018-11-11  Richard Biener  <rguenther@suse.de>
1761         * tree-vrp.h (class value_range_base): New base class for
1762         value_range containing all but the m_equiv member.
1763         (dump_value_range_base): Add.
1764         (range_includes_zero_p): Work on value_range_base.
1765         * tree-vrp.c (value_range_base::set): Split out base handling
1766         from...
1767         (value_range::set): this.
1768         (value_range::set_equiv): New.
1769         (value_range_base::value_range_base): New constructors.
1770         (value_range_base::check): Split out base handling from...
1771         (value_range::check): this.
1772         (value_range::equal_p): Refactor in terms of
1773         ignore_equivs_equal_p which is now member of the base.
1774         (value_range_base::set_undefined): New.
1775         (value_range_base::set_varying): Likewise.
1776         (value_range_base::dump):Split out base handling from...
1777         (value_range::dump): this.
1778         (value_range_base::set_and_canonicalize): Split out base handling
1779         from...
1780         (value_range::set_and_canonicalize): this.
1781         (value_range_base::union_): New.
1782         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1783         for m_vr.
1784         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1785         instead of value_range everywhere.
1786         (ipcp_vr_lattice::print): Use dump_value_range_base.
1787         (ipcp_vr_lattice::meet_with): Adjust.
1788         (ipcp_vr_lattice::meet_with_1): Likewise.
1789         (ipa_vr_operation_and_type_effects): Likewise.
1790         (propagate_vr_across_jump_function): Likewise.
1791         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1792         (ipa_get_value_range): Likewise.
1793         (ipa_set_jfunc_vr): Likewise.
1794         (ipa_compute_jump_functions_for_edge): Likewise.
1796 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
1798         PR middle-end/65703
1799         * doc/invoke.texi (Optimize Options): Add @opindex entries
1800         for the positive forms of -fno-xxx and -mno-xxx options
1801         that were lacking them.
1803 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
1805         * combine.c (make_more_copies): Only make an intermediate copy if the
1806         dest of a move is a pseudo.
1808 2018-11-09  Maya Rashish  <coypu@sdf.org>
1810         PR target/87221
1811         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1812         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1814 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1816         PR driver/41179
1817         PR middle-end/65703
1818         * doc/invoke.texi (Optimize Options): Clarify default behavior
1819         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1821 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1823         PR tree-optimization/87940
1824         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1825         Fold PLUS_EXPR correctly.
1827 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1829         PR target/87762
1830         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1831         (annotate_constant_pool_refs): Skip insns which support
1832         relative addressing.
1833         (annotate_constant_pool_refs_1): New helper function.
1834         (find_constant_pool_ref): Skip insns which support relative
1835         addression.
1836         (find_constant_pool_ref_1): New helper function.
1837         (replace_constant_pool_ref): Skip insns which support
1838         relative addressing.
1839         (replace_constant_pool_ref_1): New helper function.
1840         (s390_mainpool_start): Adapt to the new signature.
1841         (s390_mainpool_finish): Likewise.
1842         (s390_chunkify_start): Likewise.
1843         (s390_chunkify_finish): Likewise.
1844         (pass_s390_early_mach::execute): Likewise.
1845         (s390_prologue_plus_offset): Likewise.
1846         (s390_emit_prologue): Likewise.
1847         (s390_emit_epilogue): Likewise.
1849 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1851         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1852         but unsupported lastprivate with conditional modifier.
1854 2018-11-09  Jeff Law  <law@redhat.com>
1856         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1857         unused argument better.  Add gcc_unreachable to silence warning.
1859 2018-11-09  Martin Sebor  <msebor@redhat.com>
1861         PR middle-end/81824
1862         * attribs.c (has_attribute): New helper function.
1863         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1864         * attribs.h (decls_mismatched_attributes): Declare.
1865         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1866         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1867         * common.opt (-Wattribute-alias): Take an argument.
1868         (-Wno-attribute-alias): New option.
1869         * doc/extend.texi (Common Function Attributes): Document copy.
1870         (Common Variable Attributes): Same.
1871         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1872         (-Wattribute-alias): Document new option argument.
1874 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1876         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1877         matching criteria.  Remove unused array initializer.
1879 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1880             Jinsong Ji  <jji@us.ibm.com>
1882         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1883         constraints by introducing a new temporary.
1884         (_mm_cvtss_si64): Likewise.
1886 2018-11-09  Martin Liska  <mliska@suse.cz>
1888         * common.opt: Add -fipa-stack-alignment flag.
1889         * doc/invoke.texi: Document it.
1890         * final.c (rest_of_clean_state): Guard stack
1891         shrinking with flag.
1893 2018-11-09  Martin Liska  <mliska@suse.cz>
1895         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1896         to ...
1897         (ipa_discover_variable_flags): ... this.
1898         * common.opt: Come up with new flag -fipa-reference-addressable.
1899         * doc/invoke.texi: Document it.
1900         * ipa-reference.c (propagate): Call the renamed fn.
1901         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1902         Likewise.
1903         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1904         ...
1905         (ipa_discover_variable_flags): ... this.  Discover
1906         non-addressable variables only with the newly added flag.
1907         * opts.c: Enable the newly added flag with -O1 and higher
1908         optimization level.
1910 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1912         * json.cc (selftest::test_writing_literals): Fix comment.
1914 2018-11-09  Martin Liska  <mliska@suse.cz>
1916         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1917         string to a stack buffer.
1918         (aarch64_parse_cpu): Likewise.
1919         (aarch64_parse_tune): Likewise.
1921 2018-11-09  Richard Biener  <rguenther@suse.de>
1923         PR tree-optimization/87953
1924         * tree-vect-loop.c (vectorizable_reduction): For analysis
1925         always pass ops[0] to vectorizable_condition.
1927 2018-11-09  Stafford Horne  <shorne@gmail.com>
1928             Richard Henderson  <rth@twiddle.net>
1929             Joel Sherrill  <joel@rtems.org>
1931         * common/config/or1k/or1k-common.c: New file.
1932         * config/or1k/*: New.
1933         * config.gcc (or1k*-*-*): New.
1934         * configure.ac (or1k*-*-*): New test for openrisc tls.
1935         * configure: Regenerated.
1936         * doc/install.texi: Document OpenRISC triplets.
1937         * doc/invoke.texi: Document OpenRISC arguments.
1938         * doc/md.texi: Document OpenRISC.
1940 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1942         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1943         (arm7tdmi-s): Delete CPU.
1944         (arm710t): Add aliases for arm720t and arm740t.
1945         (arm720t, arm740t): Delete CPUs.
1946         (arm920t): Add aliases for arm920, arm922t and arm940t.
1947         (arm920, arm922t, arm940t): Delete CPUs.
1948         (arm10tdmi): Add alias for arm1020t.
1949         (arm1020t): Delete CPU.
1950         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1951         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1952         (arm10e): Add aliases for arm1020e and arm1022e.
1953         (arm1020e, arm1022e): Delete CPU.
1954         * config/arm/arm.md (generic_sched): Remove entries that are now
1955         handled by aliases.
1956         (generic_vfp): Likewise.
1957         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1958         changes.
1959         * config/arm/arm-tune.md: Regenerated.
1960         * config/arm/arm-tables.opt: Regenerated.
1962 2018-11-09  Richard Biener  <rguenther@suse.de>
1964         PR tree-optimization/87621
1965         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1966         op with only phi inputs.
1967         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1968         (ch_base::copy_headers): Run CSE on copied loop headers.
1969         (pass_ch_vect::process_loop_p): Simplify.
1971 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1973         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1974         for non-w64 x86_64 biarch.
1976 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1978         PR rtl-optimization/86438
1979         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1980         of in_b for the compare if in_b is SET_DEST.
1982         PR target/87793
1983         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1984         non-toplevel UNSPEC.
1986 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1988         * tree-vrp.c (value_range::check): Do not access internals
1989         directly.
1990         (value_range::singleton_p): Same.
1991         (value_range::type): Same.
1992         (vrp_finalize): Use value_range API.
1994 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1996         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1998 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2000         * vr-values.c (vr_values::get_value_range): Use value_range API
2001         instead of piecing together ranges.
2002         (vr_values::update_value_range): Same.
2004 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2006         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
2007         of performing ad-hoc calculations.
2008         * tree-ssanames.c (set_range_info): New overloaded function
2009         accepting value_range &.
2010         (get_range_info): Same.
2011         * tree-ssanames.h (set_range_info_raw): Remove.
2012         (set_range_info): New prototype.
2013         (get_range_info): Same.
2014         * tree-vrp.h (value_range::null_p): Rename to zero_p.
2015         * tree-vrp.c (value_range::null_p): Same.
2017 2018-11-09  Jan Hubicka  <jh@suse.cz>
2019         * tree.c (fld_type_variant_equal_p): Test user align flag.
2020         (flt_type_variant): Copy user align flag.
2021         (fld_incomplete_type_of): Clear it.
2023 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2025         * config/arm/neon.md (div<mode>3): New pattern.
2027 2018-11-08  Andi Kleen  <ak@linux.intel.com>
2029         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
2030         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
2031         (ix86_handle_option): Handle OPT_mptwrite.
2032         * config/i386/cpuid.h (bit_PTWRITE): Add.
2033         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
2034         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
2035         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
2036         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
2037         (ix86_option_override_internal): Handle PTA_PTWRITE.
2038         (ix86_valid_target_attribute_inner_p): Define ptwrite.
2039         (def_builtin2): Force UINT64 to be 64bit only.
2040         * config/i386/i386.h (TARGET_PTWRITE): Add.
2041         (TARGET_PTWRITE_P): Add.
2042         (PTA_PTWRITE): Add.
2043         * config/i386/i386.md: Define ptwrite.
2044         * config/i386/i386.opt: Add -mptwrite.
2045         * config/i386/immintrin.h (_ptwrite64): Add.
2046         (_ptwrite32): Add
2047         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
2048         * doc/invoke.texi: Document -mptwrite.
2050 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
2052         PR rtl-optimization/87600
2053         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
2054         * lra-constraints.c (process_alt_operands): Skip illegal hard
2055         register usage.  Prefer reloading non hard register operands.
2057 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
2059         PR other/36572
2060         * doc/invoke.texi (Optimize Options): Clarify default behavior
2061         for -fno-sched-interblock and -fno-sched-spec.
2063 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
2065         * collect2.c (linker_select):  Add USE_LLD_LD.
2066         (ld_suffixes): Add ld.lld.
2067         (main): Handle -fuse-ld=lld.
2068         * common.opt (-fuse-ld=lld): New option.
2069         * doc/invoke.texi (-fuse-ld=lld): Document.
2070         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
2072 2018-11-08  Paul Koning  <ni1d@arrl.net>
2074         * config/pdp11/constraints.md: Add "Z" series constraints for use
2075         with pre-dec and post-inc addressing.
2076         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
2077         (pdp11_expand_operands): Add int argument (word count).
2078         (pdp11_sp_frame_offset): Delete.
2079         (pdp11_cmp_length): New function.
2080         (pushpop_regeq): New function.
2081         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
2082         Add hook.
2083         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
2084         frame layout.
2085         (pdp11_initial_elimination_offset): Ditto.
2086         (pdp11_expand_operands): Add word count argument.  Bugfixes.
2087         (output_move_multiple): Change how pointer adjustment is done.
2088         (pdp11_gen_int_label): Correct format.
2089         (output_ascii): Ditto.
2090         (pdp11_asm_output_var): Add code for DEC assembler case.
2091         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
2092         value.
2093         (legitimate_const_double_p): Ditto.
2094         (pdp11_register_move_cost): Adjust for new register classes.
2095         (pdp11_regno_reg_class): Ditto.
2096         (expand_block_move): Delete.
2097         (pushpop_regeq): New function.
2098         (pdp11_legitimate_address_p): Bugfix in check for constant
2099         offset.
2100         (pdp11_sp_frame_offset): Delete.
2101         (pdp11_reg_save_size): New helper function for new frame layout.
2102         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
2103         (pdp11_expand_shift): Bugfix in check for constant shift count.
2104         (pdp11_shift_length): Ditto.
2105         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
2106         (pdp11_cmp_length): New function.
2107         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
2108         some compile options.
2109         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
2110         (CALL_USED_REGISTERS): Ditto.
2111         (ELIMINABLE_REGS): Ditto.
2112         (REGISTER_NAMES): Ditto.
2113         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
2114         constraints.
2115         (REG_CLASS_NAMES): Ditto.
2116         (REG_CLASS_CONTENTS): Ditto.  Also remove
2117         HARD_FRAME_POINTER_REGNUM.
2118         (CPU_REG_CLASS): New macro.
2119         (CLASS_MAX_NREGS): Adjust for new register classes.
2120         (FUNCTION_PROFILER): Make no-op.
2121         (may_call_alloca): Remove unused declaration.
2122         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
2123         (ASM_OUTPUT_SKIP): Fix format.
2124         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
2125         (HARD_FRAME_POINTER_REGNUM): Remove.
2126         (return): Delete.
2127         (*rts): Rename.  Remove epilogue related checks.
2128         (cmpsi, cmpdi): New insn.
2129         (cbranch<mode>4): Change to apply to SI and DI modes as well.
2130         (mov<mode>): Change constraints to enforce that push/pop
2131         destination cannot use the same register as source.
2132         (*mov<mode><cc_cc>): Ditto.
2133         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
2134         at assembly output rather than as RTL expander.
2135         (zero_extendqihi2): Bugfix in check for same registers.
2136         (adddi3_nocc): Bugfix in check for constant operand.
2137         (addsi3_nocc): Ditto.
2138         (subdi3_nocc): Ditto.
2139         (subsi3_nocc): Ditto.
2140         (negdi2_nocc): Copy input to pdp11_expand_operands.
2141         (negsi2_nocc): Ditto.
2142         (bswap2_nocc): Ditto.
2143         * config/pdp11/pdp11.opt (mlra): Fix documentation.
2144         * config/pdp11/t-pdp11: Use -Os.
2146 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
2148         * config/arm/parsecpu.awk (/alias/): New parsing rule.
2149         (/begin cpu/): Check that the cpu name hasn't been previously defined.
2150         (gen_comm_data): Print out CPU alias tables.
2151         (check_cpu): Match aliases when checking the CPU name.
2152         * config/arm/arm-protos.h (cpu_alias): New structure.
2153         (cpu_option): Add entry for aliases.
2154         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
2155         strongarm1100 and strongarm1110.
2156         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
2157         (config/arm/arm-generic.md): Remove redundant references to
2158         strongarm110, strongarm1100 and strongarm1110.
2159         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2160         Scan aliases for additional hints.
2161         (arm_parse_cpu_option_name): Also match a cpu name against the list
2162         of aliases.
2163         * config/arm/arm-tables.opt: Regenerated.
2164         * config/arm/arm-tune.md: Regenerated.
2166 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
2168         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
2169         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
2170         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2171         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
2172         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2173         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
2174         * gengtype.c (open_base_files): Add omp-general.h.
2175         * gimple.c (gimple_build_omp_critical):
2176         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
2177         gimple_omp_taskgroup_set_clauses.
2178         (gimple_build_omp_atomic_load): Add mo argument, call
2179         gimple_omp_atomic_set_memory_order.
2180         (gimple_build_omp_atomic_store): Likewise.
2181         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
2182         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
2183         instead of GSS_OMP.
2184         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
2185         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
2186         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
2187         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
2188         different value for GF_OMP_ATOMIC_NEED_VALUE.
2189         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
2190         comments.
2191         (struct gimple_statement_omp_single_layout): And remove here.
2192         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
2193         than gimple_statement_omp_single_layout.
2194         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
2195         GIMPLE_OMP_TEAMS.
2196         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
2197         (gimple_omp_subcode): Formatting fix.
2198         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
2199         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
2200         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
2201         gimple_omp_teams_host, gimple_omp_teams_set_host,
2202         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
2203         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
2204         gimple_omp_taskgroup_set_clauses): New inline functions.
2205         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
2206         (gimple_build_omp_atomic_store): Likewise.
2207         (gimple_omp_atomic_seq_cst_p): Remove.
2208         (gimple_omp_atomic_memory_order): New function.
2209         (gimple_omp_atomic_set_seq_cst): Remove.
2210         (gimple_omp_atomic_set_memory_order): New function.
2211         (gimple_build_omp_taskgroup): Add clauses argument.
2212         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
2213         (dump_gimple_omp_task): Print taskwait with depend clauses.
2214         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
2215         dump_omp_atomic_memory_order.
2216         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
2217         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
2218         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
2219         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
2220         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
2221         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
2222         ORT_UNTIED_TASKLOOP enumerators.
2223         (enum gimplify_defaultmap_kind): New.
2224         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
2225         target_map_pointers_as_0len_arrays members, add defaultmap.
2226         (new_omp_context): Initialize defaultmap member.
2227         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2228         (maybe_fold_stmt): Don't fold even in host teams regions.
2229         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
2230         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
2231         ctx->omp_firstprivatize_variable.
2232         (omp_add_variable): Don't add private/firstprivate for VLAs in
2233         ORT_TASKGROUP.
2234         (omp_default_clause): Print "taskloop" rather than "task" if
2235         ORT_*TASKLOOP.
2236         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2237         Handle new defaultmap clause kinds.
2238         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
2239         iterator to be lastprivate or private.  Fix up diagnostics if linear
2240         is used on collapse>1 simd iterator.
2241         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2242         (gimplify_omp_depend): New function.
2243         (gimplify_scan_omp_clauses): Add shared clause on parallel for
2244         combined parallel master taskloop{, simd} if taskloop has
2245         firstprivate, lastprivate or reduction clause.  Handle
2246         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
2247         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
2248         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
2249         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
2250         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
2251         lastprivate.
2252         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
2253         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
2254         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
2255         OMP_CLAUSE_{TASK,IN}_REDUCTION.
2256         (gimplify_omp_task): Handle taskwait with depend clauses.
2257         (gimplify_omp_for): Add shared clause on parallel for combined
2258         parallel master taskloop{, simd} if taskloop has firstprivate,
2259         lastprivate or reduction clause.  Use ORT_TASKLOOP or
2260         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
2261         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
2262         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
2263         __for_end and __for_range temporaries on OMP_PARALLEL for
2264         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
2265         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
2266         sandwiched in between two taskloops.
2267         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
2268         instead of ctx->omp_firstprivatize_variable.
2269         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
2270         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
2271         host teams, use gimplify_and_return_first etc. for body like
2272         for target or target data constructs, and at the end call
2273         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
2274         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
2275         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
2276         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
2277         gimple_omp_atomic_set_seq_cst calls.
2278         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
2279         case, handle taskgroup clauses.
2280         * lto-streamer-out.c (hash_tree): Handle
2281         OMP_CLAUSE_{TASK,IN}_REDUCTION.
2282         * Makefile.in (GTFILES): Add omp-general.h.
2283         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
2284         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
2285         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
2286         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
2287         BUILT_IN_GOMP_LOOP_DOACROSS_START,
2288         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
2289         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
2290         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
2291         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
2292         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
2293         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
2294         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
2295         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
2296         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
2297         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
2298         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
2299         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
2300         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
2301         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
2302         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
2303         * omp-expand.c (workshare_safe_to_combine_p): Return false for
2304         non-worksharing loops.
2305         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
2306         (determine_parallel_type): Don't combine parallel with worksharing
2307         which has _reductemp_ clause.
2308         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
2309         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
2310         if there is nonmonotonic modifier or if there is no modifier and no
2311         ordered clause.  For dynamic and guided schedule without monotonic
2312         and nonmonotonic modifier, default to nonmonotonic.
2313         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
2314         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
2315         task reductions.
2316         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
2317         there are any reduction clauses.
2318         (expand_taskwait_call): New function.
2319         (expand_teams_call): New function.
2320         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
2321         expand_teams_call for it.  Formatting fix.  Handle taskwait with
2322         depend clauses.
2323         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
2324         of worksharing loops with task reductions.
2325         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
2326         expansion of worksharing loops with task reductions.
2327         (expand_omp_sections): Handle expansion of sections with task
2328         reductions.
2329         (expand_omp_synch): For host teams call expand_omp_taskreg.
2330         (omp_memory_order_to_memmodel): New function.
2331         (expand_omp_atomic_load, expand_omp_atomic_store,
2332         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
2333         instead of gimple_omp_atomic_seq_cst_p.
2334         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
2335         depend clauses as a standalone directive.
2336         * omp-general.c (enum omp_requires): New variable.
2337         (omp_extract_for_data): Initialize have_reductemp member.  Allow
2338         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
2339         GT_EXPR loops depending on incr sign.  Formatting fixes.
2340         * omp-general.h (struct omp_for_data): Add have_reductemp member.
2341         (enum omp_requires): New enum.
2342         (omp_requires_mask): Declare.
2343         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
2344         Fix comment typos.
2345         * omp-low.c (struct omp_context): Add task_reductions and
2346         task_reduction_map fields.
2347         (is_host_teams_ctx): New function.
2348         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
2349         (use_pointer_for_field): Use is_global_var instead of
2350         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
2351         in outer contexts.
2352         (build_outer_var_ref): Ignore taskgroup outer contexts.
2353         (delete_omp_context): Release task_reductions and task_reduction_map.
2354         (scan_sharing_clauses): Don't add any fields for reduction clause on
2355         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
2356         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
2357         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
2358         Handle OMP_CLAUSE_NONTEMPORAL.
2359         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
2360         needed.
2361         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
2362         clauses with task modifier.
2363         (scan_omp_task): Handle taskwait with depend clauses.
2364         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
2365         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
2366         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
2367         (scan_omp_for): Fix comment formatting.
2368         (scan_omp_teams): Handle host teams constructs.
2369         (check_omp_nesting_restrictions): Allow teams with no outer
2370         OpenMP context.  Adjust diagnostics for teams strictly nested into
2371         some explicit OpenMP construct other than target.  Allow OpenMP atomics
2372         inside of simd regions.
2373         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
2374         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
2375         taskreg_nesting_level while scanning host teams construct.
2376         (task_reduction_read): New function.
2377         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
2378         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
2379         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
2380         second argument create_tmp_var if it is NULL.  Don't ignore shared
2381         clauses in is_host_teams_ctx contexts.  Handle
2382         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
2383         clauses.
2384         (lower_reduction_clauses): Ignore reduction clauses with task
2385         modifier.  Remove second argument create_tmp_var if it is NULL.
2386         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
2387         (lower_send_clauses): Ignore reduction clauses with task modifier.
2388         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
2389         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
2390         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
2391         rely that it is the last stmt in body so far.  Ignore outer taskgroup
2392         contexts.
2393         (omp_task_reductions_find_first, omp_task_reduction_iterate,
2394         lower_omp_task_reductions): New functions.
2395         (lower_omp_sections): Handle reduction clauses with taskgroup
2396         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
2397         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
2398         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
2399         modifiers.
2400         (lower_omp_taskgroup): Handle taskgroup reductions.
2401         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
2402         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
2403         (lower_depend_clauses): If there are any
2404         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
2405         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
2406         seen, assume lowering is done already and return early.  Set kind
2407         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
2408         (lower_omp_taskreg): Handle reduction clauses with task modifier on
2409         parallel construct.  Handle reduction clause on taskloop construct.
2410         Handle taskwait with depend clauses.
2411         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
2412         for host teams constructs.
2413         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
2414         nontemporal and _reductemp_ clause entries.
2415         (omp_clause_code_name): Likewise.
2416         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
2417         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
2418         * tree-core.h (enum omp_clause_code): Add
2419         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
2420         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
2421         (struct tree_base): Add omp_atomic_memory_order field into union.
2422         Remove OMP_ATOMIC_SEQ_CST comment.
2423         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
2424         and OMP_CLAUSE_DEPEND_DEPOBJ.
2425         (struct tree_omp_clause): Add subcode.defaultmap_kind.
2426         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
2427         OpenMP constructs with body and clauses operands.
2428         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
2429         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
2430         (OMP_TASKGROUP_CLAUSES): Define.
2431         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
2432         OMP_CLAUSE__LOOPTEMP_.
2433         (OMP_ATOMIC_SEQ_CST): Remove.
2434         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
2435         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
2436         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
2437         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
2438         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
2439         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
2440         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
2441         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
2442         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
2443         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
2444         Define.
2445         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
2446         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2447         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
2448         (convert_local_omp_clauses): Likewise.  Remove useless test.
2449         * tree-parloops.c (create_call_for_reduction_1): Pass
2450         OMP_MEMORY_ORDER_RELAXED as new argument to
2451         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
2452         * tree-pretty-print.c (dump_omp_iterators): New function.
2453         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
2454         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
2455         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
2456         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
2457         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
2458         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
2459         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
2460         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
2461         (dump_omp_atomic_memory_order): New function.
2462         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
2463         taskwait with depend clauses.
2464         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
2465         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
2466         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
2467         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
2468         write_ts_omp_clause_tree_pointers): Likewise.
2470 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2472         PR ipa/86395
2473         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
2474         "-missed", "-note", and "-all" sub-options.
2475         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
2476         API.
2477         (can_early_inline_edge_p): Likewise.
2478         (want_early_inline_function_p): Likewise.
2479         (want_inline_self_recursive_call_p): Likewise.
2480         (recursive_inlining): Likewise.
2481         (inline_small_functions): Likewise.
2482         (flatten_function): Likewise.
2483         (ipa_inline): Likewise.
2484         (inline_always_inline_functions): Likewise.
2485         (early_inline_small_functions): Likewise.
2486         (early_inliner): Likewise.
2487         * tree-inline.c (expand_call_inline): Likewise.
2489 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2491         * pretty-print.c (pp_format): Handle %f.
2492         (selftest::test_pp_format): Add test of %f.
2493         * pretty-print.h (pp_double): New macro.
2495 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2497         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
2498         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
2499         (dump_pretty_printer::decode_format): Implement "%C" for
2500         cgraph_node *.
2501         (selftest::test_capture_of_dump_calls): Rename "where" to
2502         "stmt_loc".  Convert test_decl to a function decl and set its
2503         location.  Add a symbol_table_test RAII instance and a
2504         cgraph_node, using it to test "%C" and dump_symtab_node.
2506 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
2508         PR middle-end/87916
2509         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
2511 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2513         * cgraph.c: Include "selftest.h".
2514         (saved_symtab): New variable.
2515         (selftest::symbol_table_test::symbol_table_test): New ctor.
2516         (selftest::symbol_table_test::~symbol_table_test): New dtor.
2517         (selftest::test_symbol_table_test): New test.
2518         (selftest::cgraph_c_tests): New.
2519         * cgraph.h (saved_symtab): New decl.
2520         (selftest::symbol_table_test): New class.
2521         * selftest-run-tests.c (selftest::run_tests): Call
2522         selftest::cgraph_c_tests.
2523         * selftest.h (selftest::cgraph_c_tests): New decl.
2525 2018-11-08  Richard Biener  <rguenther@suse.de>
2527         * tree-data-ref.h (lambda_int): New typedef.
2528         (lambda_vector_gcd): Adjust.
2529         (lambda_vector_new): Likewise.
2530         (lambda_matrix_new): Likewise.
2531         * tree-data-ref.c  (print_lambda_vector): Adjust.
2533 2018-11-08  Richard Biener  <rguenther@suse.de>
2535         PR tree-optimization/87929
2536         * tree-complex.c (expand_complex_comparison): Clean EH.
2538 2018-11-08  Martin Liska  <mliska@suse.cz>
2540         * doc/extend.texi: Reword.
2541         * predict.c (expr_expected_value_1): Likewise.
2543 2018-11-08  Richard Biener  <rguenther@suse.de>
2545         PR tree-optimization/87913
2546         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
2547         of extreme values to ordered comparisons.
2549 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
2551         PR middle-end/42726
2552         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
2554 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
2556         * doc/invoke.texi: Remove leading dash from @opindex entries
2557         throughout the file.
2559 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
2561         PR driver/80828
2562         * doc/invoke.texi (Option Summary): Add -e and --entry.
2563         (Link Options): Likewise.
2565 2018-11-07  Nathan Sidwell  <nathan@acm.org>
2567         PR 87926
2568         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
2569         --disable-checking bootstrap.
2571 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
2573         * configure: Regenerated.
2575 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2577         PR c/87691
2578         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
2579         to the mode of the widest field iff the widest field has mode class
2580         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
2581         reference.
2583 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
2585         * common.opt: Add -Wattribute-warning.
2586         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
2587         * expr.c (expand_expr_real_1): Add new attribute to warning_at
2588         call to allow user configure behavior of "warning" attribute.
2590 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
2592         * target.def: Put @: after every vs., e.g., and i.e. where it is
2593         followed by whitespace.
2594         * doc/extend.texi: Ditto.
2595         * doc/fragments.texi: Ditto.
2596         * doc/gimple.texi: Ditto.
2597         * doc/implement-c.texi: Ditto.
2598         * doc/install.texi: Ditto.
2599         * doc/invoke.texi: Ditto.
2600         * doc/md.texi: Ditto.
2601         * doc/plugins.texi: Ditto.
2602         * doc/rtl.texi: Ditto.
2603         * doc/sourcebuild.texi: Ditto.
2604         * doc/tm.texi.in: Ditto.
2605         * doc/ux.texi: Ditto.
2606         * doc/tm.texi: Regenerate.
2608 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2610         * config/arm/arm-cpus.in (ares): New entry.
2611         * config/arm/arm-tables.opt: Regenerate.
2612         * config/arm/arm-tune.md: Likewise.
2613         * doc/invoke.texi (ARM Options): Document ares.
2615 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2617         * config/aarch64/aarch64-cores.def (ares): Define.
2618         * config/aarch64/aarch64-tune.md: Regenerate.
2619         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
2621 2018-11-07  Jan Hubicka  <jh@suse.cz>
2623         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
2624         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
2625         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
2626         TYPE_DECL.
2628 2018-11-07  Richard Biener  <rguenther@suse.de>
2630         PR tree-optimization/87914
2631         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
2632         of nested cycles.
2633         (vectorizable_reduction): Handle shifts and rotates by dispatching
2634         to vectorizable_shift.
2635         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
2636         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
2637         def cases.
2638         (vectorizable_shift): Export and handle being called as
2639         vect_nested_cycle.
2640         (vect_analyze_stmt): Call vectorizable_shift after
2641         vectorizable_reduction.
2642         * tree-vectorizer.h (vectorizable_shift): Declare.
2644 2018-11-07  Jan Hubicka  <jh@suse.cz>
2646         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
2647         than const decls in TREE_VALUE of enum.
2648         (dump_type_inheritance_graph): Improve duplicate dumping.
2649         (free_enum_values): New.
2650         (build_type_inheritance_graph): Use it.
2651         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
2652         which are not main variants or not ODR types.
2653         (verify_type_variant): Expect variants to have no TYPE_VALUES.
2655 2018-11-07  Richard Biener  <rguenther@suse.de>
2657         * ipa-inline.c (want_inline_small_function_p): Compute
2658         big_speedup_p lazily and last.
2660 2018-11-07  Jan Hubicka  <jh@suse.cz>
2662         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
2663         building incomplete variant of complete type.
2664         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
2665         variant of complete type.
2667 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2669         * config/mips/mips.c: Fix typo in documentation of
2670         mips_loongson_ext2_prefetch_cookie.
2671         (mips_option_override): fix brain twister logical.
2672         * config/mips/mips.h: Fix typo in documentation of
2673         ISA_HAS_CTZ_CTO and define pattern.
2674         * config/mips/mips.md (prefetch): Hoist EXT2 above
2675         the 2EF/EXT block.
2676         (prefetch_indexed): Hoist EXT2 above the EXT block.
2678 2018-11-07  Jan Hubicka  <jh@suse.cz>
2680         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
2681         return and parameter types of function and method types.
2682         (free_lang_data_in_cgraph): Update.
2684 2018-11-07  Martin Liska  <mliska@suse.cz>
2686         PR rtl-optimization/87868
2687         * postreload-gcse.c (eliminate_partially_redundant_load): Set
2688         threshold to max_count if we would overflow.
2689         * profile-count.h: Make max_count a public constant.
2691 2018-11-07  Martin Liska  <mliska@suse.cz>
2693         * mem-stats.h: Fix GNU coding style.
2695 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2697         * config/mips/gs264e.md: New.
2698         * config/mips/mips-cpus.def: Define gs264e.
2699         * config/mips/mips-tables.opt: Regenerate.
2700         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2701         gs264e.
2702         (mips_issue_rate): Add support for gs264e.
2703         (mips_multipass_dfa_lookahead): Likewise.
2704         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2705         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2706         (MIPS_ASE_MSA_SPEC): New.
2707         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2708         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2709         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2710         * config/mips/mips.md: Include gs264e.md.
2711         (processor): Add gs264e.
2712         * config/mips/mips.opt (MSA): Use Mask instead of Var.
2713         * doc/invoke.texi: Add gs264e to supported architectures.
2715 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2717         * config/mips/gs464e.md: New.
2718         * config/mips/mips-cpus.def: Define gs464e.
2719         * config/mips/mips-tables.opt: Regenerate.
2720         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2721         gs464e.
2722         (mips_issue_rate): Add support for gs464e.
2723         (mips_multipass_dfa_lookahead): Likewise.
2724         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2725         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2726         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2727         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2728         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2729         * config/mips/mips.md: Include gs464e.md.
2730         (processor): Add gs464e.
2731         * doc/invoke.texi: Add gs464e to supported architectures.
2733 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2735         * config/mips/loongson3a.md: Rename to ...
2736         * config/mips/gs464.md: ... here.
2737         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2738         as an alias of gs464 processor.
2739         * config/mips/mips-tables.opt: Regenerate.
2740         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2741         instead of PROCESSOR_LOONGSON_3A.
2742         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2743         TUNE_LOONGSON_3A.
2744         (mips_option_override): Enable MMI and EXT for gs464.
2745         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2746         Rename TUNE_LOONGSON_3A to TUNE_GS464.
2747         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2748         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2749         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2750         TARGET_LOONGSON_3A.
2751         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2752         (processor): Add gs464;
2753         * doc/invoke.texi: Add gs464 to supported architectures.
2755 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2757         * config/mips/mips-protos.h
2758         (mips_loongson_ext2_prefetch_cookie): New prototype.
2759         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2760         (mips_option_override): Enable TARGET_LOONGSON_EXT when
2761         TARGET_LOONGSON_EXT2 is true.
2762         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2763         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2764         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2765         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2766         TARGET_LOONGSON_EXT2.
2767         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2768         (define_insn "ctz<mode>2"): New insn pattern.
2769         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2770         (define_insn "prefetch_indexed_<mode>"): Include
2771         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2772         * config/mips/mips.opt (-mloongson-ext2): Add option.
2773         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2775 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2777         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2778         __mips_loongson_ext.
2779         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2780         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2781         -mloongson-ext.
2782         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2783         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2784         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2785         instead of TARGET_LOONGSON_3A.
2786         * config/mips/mips.opt (-mloongson-ext): Add option.
2787         * gcc/doc/invoke.texi (-mloongson-ext): Document.
2789 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2791         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2792         * config/mips/loongson.md: Move to ...
2793         * config/mips/loongson-mmi.md: here; Adjustment.
2794         * config/mips/loongson.h: Move to ...
2795         State as deprecated. Include loongson-mmiintrin.h for back
2796         compatibility and warning.
2797         * config/mips/loongson-mmiintrin.h: ... here.
2798         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2799         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2800         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2801         (mips_option_override): Make sure MMI use hard float;
2802         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2803         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2804         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2805         TARGET_LOONGSON_VECTORS.
2806         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2807         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2808         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2809         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2810         -mloongson-mmi.
2811         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2812         TARGET_LOONGSON_VECTORS.
2813         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2814         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2815         (Loongson MMI patterns): Include loongson-mmi.md instead of
2816         loongson.md.
2817         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2818         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2820 2018-11-07  Richard Biener  <rguenther@suse.de>
2822         PR lto/87906
2823         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2824         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2826 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2828         PR rtl-optimization/87874
2829         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2830         const wide ints.
2832 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2834         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2835         if not in indexed or indirect form.
2836         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2837         (bswap<mode>2_store): Ditto.
2839 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2841         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2842         the UNSPEC.
2844 2018-11-06  Richard Biener  <rguenther@suse.de>
2846         PR tree-optimization/86850
2847         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2848         instead of src.m_vec.
2850 2018-11-06  Jan Hubicka  <jh@suse.cz>
2852         * tree.c (fld_simplified_type_name): Break out form ...
2853         (free_lang_data_in_type): ... here.
2854         (fld_type_variant_equal_p): Use it.
2856 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2858         * config/default-d.c: Include memmodel.h.
2860         * config/sol2-d.c: New file.
2861         * config/t-sol2 (sol2-d.o): New rule.
2862         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2863         target_has_targetdm.
2865 2018-11-06  Jan Hubicka  <jh@suse.cz>
2867         * tree.c (fld_type_variant): Also copy alignment; be sure that
2868         new variant is equal.
2870 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2872         PR target/87762
2873         * config/s390/s390.md: Add relative_long attribute.
2875 2018-11-06  Jan Hubicka  <jh@suse.cz>
2877         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2878         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2879         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2880         stream TYPE_NEEDS_CONSTRUCTING.
2881         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2882         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2884 2018-11-06  Richard Biener  <rguenther@suse.de>
2886         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2887         dump-scope ...
2888         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2890 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2892         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2893         REG_CLASS_CONTENTS[GEN_REGS].
2894         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2896 2018-11-06  Jan Hubicka  <jh@suse.cz>
2898          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2900 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2902         PR sanitizer/80953
2903         * config/sol2.h (ASAN_CC1_SPEC): Define.
2904         (LD_WHOLE_ARCHIVE_OPTION): Define.
2905         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2906         (ASAN_REJECT_SPEC): Provide default.
2907         (LIBASAN_EARLY_SPEC): Define.
2908         (LIBTSAN_EARLY_SPEC): Define.
2909         (LIBLSAN_EARLY_SPEC): Define.
2910         * config/i386/sol2.h (CC1_SPEC): Redefine.
2911         (ASAN_REJECT_SPEC): Define.
2913         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2914         (TARGET_ASAN_SHADOW_OFFSET): Define.
2915         (sparc_asan_shadow_offset): New function.
2916         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2917         (ASAN_REJECT_SPEC): Define.
2919 2018-11-06  Jan Hubicka  <jh@suse.cz>
2921         * tree.c (fld_type_variant): Copy canonical type.
2922         (fld_incomplete_type_of): Check that canonical types looks sane;
2923         copy canonical type.
2924         (verify_type): Accept when incomplete type has complete canonical type.
2926 2018-11-06  Jan Hubicka  <jh@suse.cz>
2928         * tree.c (free_lang_data): Reset overwite_assembler_name,
2929         print_xnode, print_decl, print_type and print_identifier of
2930         langhooks.
2932 2018-11-06  Richard Biener  <rguenther@suse.de>
2934         PR tree-optimization/87889
2935         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2936         Do nothing if old and new arg are the same
2938 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2940         PR target/87723
2941         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2942         attributes for operands 3 and 4.
2944 2018-11-06  Richard Biener  <rguenther@suse.de>
2946         PR middle-end/18041
2947         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2948         matching bitfield insertion.
2950 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2952         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2953         (reg_next_debug_use): New.
2954         (attempt_change): Propagate adjusted expression into affected
2955         debug insns.
2956         (merge_in_block): Track uses in debug insns.
2957         (pass_inc_dec::execute): Allocate and release
2958         reg_next_debug_use.
2960 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2962         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2963         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2964         (_mm512_maskz_fixupimm_round_pd): Ditto.
2965         (_mm512_fixupimm_round_ps): Ditto.
2966         (_mm512_maskz_fixupimm_round_ps): Ditto.
2967         (_mm_fixupimm_round_sd): Ditto.
2968         (_mm_maskz_fixupimm_round_sd): Ditto.
2969         (_mm_fixupimm_round_ss): Ditto.
2970         (_mm_maskz_fixupimm_round_ss): Ditto.
2971         (_mm512_fixupimm_pd): Ditto.
2972         (_mm512_maskz_fixupimm_pd): Ditto.
2973         (_mm512_fixupimm_ps): Ditto.
2974         (_mm512_maskz_fixupimm_ps): Ditto.
2975         (_mm_fixupimm_sd): Ditto.
2976         (_mm_maskz_fixupimm_sd): Ditto.
2977         (_mm_fixupimm_ss): Ditto.
2978         (_mm_maskz_fixupimm_ss): Ditto.
2979         (_mm512_mask_fixupimm_round_pd): Update builtin.
2980         (_mm512_mask_fixupimm_round_ps): Ditto.
2981         (_mm_mask_fixupimm_round_sd): Ditto.
2982         (_mm_mask_fixupimm_round_ss): Ditto.
2983         (_mm512_mask_fixupimm_pd): Ditto.
2984         (_mm512_mask_fixupimm_ps): Ditto.
2985         (_mm_mask_fixupimm_sd): Ditto.
2986         (_mm_mask_fixupimm_ss): Ditto.
2987         * config/i386/avx512vlintrin.h:
2988         (_mm256_fixupimm_pd): Update parameters and builtin.
2989         (_mm256_maskz_fixupimm_pd): Ditto.
2990         (_mm256_fixupimm_ps): Ditto.
2991         (_mm256_maskz_fixupimm_ps): Ditto.
2992         (_mm_fixupimm_pd): Ditto.
2993         (_mm_maskz_fixupimm_pd): Ditto.
2994         (_mm_fixupimm_ps): Ditto.
2995         (_mm_maskz_fixupimm_ps): Ditto.
2996         (_mm256_mask_fixupimm_pd): Update builtin.
2997         (_mm256_mask_fixupimm_ps): Ditto.
2998         (_mm_mask_fixupimm_pd): Ditto.
2999         (_mm_mask_fixupimm_ps): Ditto.
3000         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
3001         * config/i386/i386-builtin.def: Update builtin definitions.
3002         * config/i386/i386.c: Handle new builtin types and remove useless ones.
3003         * config/i386/sse.md: Update VFIXUPIMM* patterns.
3004         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3005         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3006         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
3007         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3008         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3009         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
3010         * config/i386/subst.md:
3011         (round_saeonly_sd_mask_operand4): Add new subst_attr.
3012         (round_saeonly_sd_mask_op4): Ditto.
3013         (round_saeonly_expand_operand5): Ditto.
3014         (round_saeonly_expand): Update.
3016 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
3018         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
3020 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
3022         PR rtl-optimization/87871
3023         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
3025 2018-11-05  Paul Koning  <ni1d@arrl.net>
3027         * doc/sourcebuild.texi (target attributes): Document new "inf"
3028         effective target keyword.
3030 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3032         * config/s390/s390.c (s390_register_move_cost): Increase costs for
3033         moves involving the CC reg.
3035 2018-11-05  Richard Biener  <rguenther@suse.de>
3037         PR tree-optimization/87873
3038         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
3039         argument.
3040         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
3041         * tree-vect-loop.c (vect_transform_loop): When splitting the
3042         loop exit also create forwarder PHIs for constants.
3043         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
3044         Handle constant to_arg, add extra checking we match up the correct
3045         PHIs.
3047 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3049         * config/s390/s390.md: QImode and HImode for load on condition.
3051 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3053         * config/s390/predicates.md: Fix typo.
3054         * config/s390/s390.md: Allow immediates for load on condition.
3056 2018-11-05  Martin Liska  <mliska@suse.cz>
3058         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
3059         * bitmap.h (struct bitmap_usage): Likewise.
3060         * ggc-common.c (SCALE): Remove.
3061         (LABEL): Likewise.
3062         (struct ggc_usage): Use SIZE_AMOUNT. And update
3063         compare method.
3064         * ggc-page.c (SCALE): Remove.
3065         (STAT_LABEL): Remove.
3066         (ggc_print_statistics): Use SIZE_AMOUNT.
3067         * gimple.h (SCALE): Remove.
3068         (LABEL): Likewise.
3069         * input.c (ONE_K): Remove.
3070         (ONE_M): Likewise.
3071         (SCALE): Likewise.
3072         (STAT_LABEL): Likewise.
3073         (FORMAT_AMOUNT): Likewise.
3074         (dump_line_table_statistics): Use SIZE_AMOUNT.
3075         * mem-stats.h (struct mem_usage): Likewise.
3076         * rtl.c (dump_rtx_statistics): Likewise.
3077         (rtx_alloc_counts): Change type to size_t.
3078         (rtx_alloc_sizes): Likewise.
3079         (rtx_count_cmp): New.
3080         (dump_rtx_statistics): Sort first based on counts.
3081         * tree.c (tree_nodes_cmp): New.
3082         (tree_codes_cmp): New.
3083         (dump_tree_statistics): Sort first based on counts.
3084         * system.h (ONE_K): New.
3085         (ONE_M): Likewise.
3086         (SIZE_SCALE): Likewise.
3087         (SIZE_LABEL): Likewise.
3088         (SIZE_AMOUNT): Likewise.
3089         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
3090         * tree-dfa.c (dump_dfa_stats): Likewise.
3091         * tree-phinodes.c (phinodes_print_statistics): Likewise.
3092         * tree-ssanames.c (ssanames_print_statistics): Likewise.
3093         * tree.c (dump_tree_statistics): Likewise.
3094         * vec.c (struct vec_usage): Likewise.
3095         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
3096         trigger a -Werror=format-overflow with
3097         --enable-gather-detailed-stats.
3099 2018-11-05  Martin Liska  <mliska@suse.cz>
3101         * mem-stats.h (mem_alloc_description::release_instance_overhead):
3102         Return T *.
3103         * vec.c (struct vec_usage): Register m_element_size.
3104         (vec_prefix::register_overhead): New arguments: elements and
3105         element_size.
3106         (vec_prefix::release_overhead): Subtract elements.
3107         * vec.h (struct vec_prefix): Change signature.
3108         (va_heap::reserve): Pass proper arguments.
3109         (va_heap::release): Likewise.
3111 2018-11-05  Martin Liska  <mliska@suse.cz>
3113         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
3114         style.
3115         * vec.c: Likewise.
3117 2018-11-05  Richard Biener  <rguenther@suse.de>
3119         * tree-scalar-evolution.h (final_value_replacement_loop): Update
3120         prototype.
3121         * tree-scalar-evolution.c (final_value_replacement_loop): Return
3122         whether anything was done.
3123         (scev_const_prop): Remove constant propagation part, fold
3124         remains into ...
3125         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
3126         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
3127         conditionally.
3129 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
3131         PR tree-optimization/87859
3132         * gimple-ssa-store-merging.c (struct merged_store_group): Add
3133         only_constants and first_nonmergeable_order members.
3134         (merged_store_group::merged_store_group): Initialize them.
3135         (merged_store_group::do_merge): Clear only_constants member if
3136         adding something other than INTEGER_CST store.
3137         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
3138         stores with order >= first_nonmergeable_order.  Use
3139         merged_store->only_constants instead of always recomputing it.
3140         Set merged_store->first_nonmergeable_order if we've skipped any
3141         stores.  Attempt to merge overlapping INTEGER_CST stores that
3142         we would otherwise skip.
3144         PR sanitizer/87837
3145         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
3147 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
3149         PR target/87853
3150         * config/i386/emmintrin.h (__v16qs): New to cope with option
3151         -funsigned-char.
3152         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
3153         (_mm_cmplt_epi8): Likewise.
3154         (_mm_cmpgt_epi8): Likewise.
3156 2018-11-05  Richard Biener  <rguenther@suse.de>
3158         PR rtl-optimization/87852
3159         * fwprop.c (use_killed_between): Only consider single-defs of the
3160         use whose definition statement dominates the use.
3162 2018-11-05  Martin Liska  <mliska@suse.cz>
3164         PR web/87829
3165         * doc/invoke.texi: Remove options that are
3166         not disabled with -Os.
3168 2018-11-05  Martin Liska  <mliska@suse.cz>
3170         PR c/87811
3171         * doc/extend.texi: Update constrain about the last argument
3172         of __builtin_expect_with_probability.
3174 2018-11-05  Martin Liska  <mliska@suse.cz>
3176         PR c/87811
3177         * predict.c (expr_expected_value_1): Verify
3178         that last argument is a real constants and emit
3179         error.
3181 2018-11-05  Martin Liska  <mliska@suse.cz>
3183         PR gcov-profile/77698
3184         * ipa-profile.c (ipa_profile): Adjust hotness threshold
3185         only in LTO mode.
3187 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3189         PR tree-optimization/86572
3190         * builtins.c (c_strlen): Handle negative offsets in a safe way.
3192 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3194         PR tree-optimization/87672
3195         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
3196         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
3198 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
3200         PR middle-end/58372
3201         * cfgexpand.c (pass_expand::execute): Move the call to
3202         finish_eh_generation in front of the call to expand_stack_alignment.
3204 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
3206         * common/config/i386/i386-common.c (processor_alias_table): Add
3207         znver2 entry.
3208         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
3209         (case ${target}): Add znver2.
3210         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
3211         -march=native recognize znver2 processors.
3212         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
3213         * config/i386/i386.c (m_znver2): New definition.
3214         (m_ZNVER): New definition.
3215         (m_AMD_MULTIPLE): Includes m_znver2.
3216         (processor_cost_table): Add znver2 entry.
3217         (processor_target_table): Add znver2 entry.
3218         (get_builtin_code_for_version): Set priority for
3219         PROCESSOR_ZNVER2.
3220         (processor_model): Add M_AMDFAM17H_ZNVER2.
3221         (arch_names_table): Ditto.
3222         (ix86_reassociation_width): Include znver2.
3223         * config/i386/i386.h (TARGET_znver2): New definition.
3224         (struct ix86_size_cost): Add TARGET_ZNVER2.
3225         (enum processor_type): Add PROCESSOR_ZNVER2.
3226         * config/i386/i386.md (define_attr "cpu"): Add znver2.
3227         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
3228         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
3229         (ix86_adjust_cost): Add znver2.
3230         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
3231         * gcc/doc/extend.texi: Add details about znver2.
3232         * gcc/doc/invoke.texi: Add details about znver2.
3234 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
3236         PR target/87079
3238         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
3239         pattern.
3241 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3243         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
3244         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
3246 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
3248         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
3249         shrink-wrapping when -mtrack-speculation.
3251 2018-11-02  Richard Biener  <rguenther@suse.de>
3253         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
3254         (pop_cost_one_pair): Do not free pair.
3255         (pop_best_coalesce): Likewise.
3256         (create_coalesce_list): Initialize obstack.
3257         (delete_coalesce_list): Free obstack.
3258         (find_coalesce_pair): Obstack-allocate coalesce pairs.
3259         (add_cost_one_coalesce): Likewise.
3260         (struct live_track): Remove bitmap pointer indirections.
3261         (new_live_track): Adjust.
3262         (delete_live_track): Likewise.
3263         (live_track_remove_partition): Likewise.
3264         (live_track_add_partition): Likewise.
3265         (live_track_live_p): Likewise.
3266         (live_track_process_def): Likewise.
3267         (live_track_clear_base_vars): Likewise.
3269 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3271         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
3272         * configure: Regenerate.
3273         * config.in: Regenerate.
3274         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
3275         HAVE_AS_REGISTER_PSEUDO_OP guard.
3276         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
3277         !HAVE_AS_REGISTER_PSEUDO_OP support.
3278         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
3279         guard.
3281 2018-11-02  Richard Biener  <rguenther@suse.de>
3283         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
3284         with is_type_die.
3286 2018-11-02  Richard Biener  <rguenther@suse.de>
3288         PR tree-optimization/87776
3289         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
3290         executable when iterating but running into rpo-vn-max-loop-depth
3291         and not eliding the iteration.
3293 2018-11-30  Jan Hubicka  <jh@suse.cz>
3295         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
3296         TYPE_DECL.
3298 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
3300         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
3301         prototype.
3302         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
3303         Combine with rs6000_address_for_fpconvert.
3304         (rs6000_address_for_fpconvert): Combine with
3305         rs6000_force_indexed_or_indirect_mem.
3306         (rs6000_expand_vector_init): Change function call from
3307         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
3308         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
3309         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
3310         (floatsi<mode>2_lfiwax_mem): Ditto.
3311         (floatunssi<mode>2_lfiwzx): Ditto.
3312         (floatunssi<mode>2_lfiwzx_mem): Ditto.
3313         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
3314         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
3315         (fix_trunc<mode>si2_stfiwx): Ditto.
3316         (fixuns_trunc<mode>si2_stfiwx): Ditto.
3317         (float_<mode>si2_hw): Ditto.
3318         (floatuns_<mode>si2_hw): Ditto.
3319         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
3320         (vsx_splat_<mode>): Ditto.
3322 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
3324         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
3325         at end of assembler input text.
3326         * configure: Regenerate.
3328 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
3330         PR tree-optimization/87826
3331         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
3332         negative or larger or equal to type's precision.
3334 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
3336         * opts.c (default_options_table): Do not enable
3337         OPT_fdelayed_branch at -Og.
3338         * doc/invoke.texi (-fdelayed-branch): Document it.
3340 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
3342         * optabs-libfuncs.c (build_libfunc_function_visibility):
3343         New, split out from...
3344         (build_libfunc_function): ... here.
3345         (init_one_libfunc_visibility): New, split out from ...
3346         (init_one_libfunc): ... here.
3348         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
3349         scratch register need not be early-clobber.  Document the reason
3350         why we cannot use ST<OP>.
3352 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
3354         PR bootstrap/82856
3355         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
3356         line for second argument of AC_DEFINE_UNQUOTED.
3357         * doc/install.texi (Tools/packages necessary for modifying GCC):
3358         Update to autoconf 2.69 and automake 1.15.1.
3359         * aclocal.m4, config.in, configure: Regenerate.
3361 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
3363         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
3364         initialization.
3366 2018-10-31  Martin Liska  <mliska@suse.cz>
3368         PR driver/83193
3369         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
3370         Add new argument invalid_extension.
3371         (aarch64_get_all_extension_candidates): New function.
3372         (aarch64_rewrite_selected_cpu): Add NULL to function call.
3373         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
3374         new argument.
3375         (aarch64_get_all_extension_candidates): New function.
3376         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
3377         argument invalid_extension.
3378         (aarch64_parse_cpu): Likewise.
3379         (aarch64_print_hint_for_extensions): New function.
3380         (aarch64_validate_mcpu): Provide hint about invalid extension.
3381         (aarch64_validate_march): Likewise.
3382         (aarch64_handle_attr_arch): Pass new argument.
3383         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
3384         (aarch64_handle_attr_isa_flags): Likewise.
3386 2018-10-31  Richard Biener  <rguenther@suse.de>
3388         PR middle-end/70359
3389         PR middle-end/86270
3390         * tree-outof-ssa.c (insert_backedge_copies): Restrict
3391         copy generation to useful cases.  Place the copy before
3392         the definition of the backedge value when possible.
3394 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3396         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
3397         * config/arc/arc.c (arc_active_insn): New function.
3398         (check_store_cacheline_hazard): Likewise.
3399         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
3400         (arc_override_options): Disable delay slot scheduler for older
3401         A7.
3402         (arc_store_addr_hazard_p): New implementation, old one renamed to
3403         ...
3404         (arc_store_addr_hazard_internal_p): Renamed.
3405         (arc_reorg): Don't combine into brcc instructions which are part
3406         of hardware hazard solution.
3407         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
3408         (tune_arc700): Likewise.
3409         * config/arc/arc.opt (arc7xx): New tune value.
3410         * config/arc/arc700.md: Improve A7 scheduler.
3412 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3414         * config/arc/arc.c (arc_override_options): Remove
3415         TARGET_COMPACT_CASESI.
3416         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
3417         (CASE_VECTOR_MODE): Likewise.
3418         (CASE_VECTOR_PC_RELATIVE): Likewise.
3419         (CASE_VECTOR_SHORTEN_MODE): Likewise.
3420         (CASE_VECTOR_SHORTEN_MODE1): Delete.
3421         (ADDR_VEC_ALIGN): Update.
3422         (ASM_OUTPUT_CASE_LABEL): Undefine.
3423         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
3424         (TARGET_BI_BIH): Define.
3425         (DEFAULT_BRANCH_INDEX): Likewise.
3426         * config/arc/arc.md (casesi): Rework to accept BI/BIH
3427         instructions, remove compact_casesi use case.
3428         (casesi_compact_jump): Remove.
3429         (casesi_dispatch): New pattern.
3430         * config/arc/arc.opt: Add mbranch-index option. Deprecate
3431         compact_casesi option.
3432         * doc/invoke.texi: Document mbranch-index option.
3434 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3436         * config/arc/arc.c (arc_get_tp): Remove function.
3437         (arc_emit_call_tls_get_addr): Likewise.
3438         (arc_call_tls_get_addr): New function.
3439         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
3440         * config/arc/arc.md (tls_load_tp_soft): Remove.
3441         (tls_gd_get_addr): Likewise.
3443 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3445         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
3446         (mulsi3_600_lib): Remove pattern.
3447         (umulsi3_highpart_600_lib_le): Likewise.
3448         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
3449         (umulsidi3): Remove call to umulsidi3_600_lib.
3450         (umulsidi3_600_lib): Remove pattern.
3451         (peephole2): Remove peephole using the above deprecated patterns.
3453 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
3455         PR target/87374
3456         * config/arm/arm.c (arm_option_check_internal): Disable the combined
3457         use of -mslow-flash-data and -mword-relocations.
3458         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
3459         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
3460         flag_pic.
3461         * doc/invoke.texi (-mword-relocations): Mention conflict with
3462         -mslow-flash-data.
3463         (-mslow-flash-data): Reciprocally.
3465 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
3467         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
3468         16-byte modes held in GP registers to use an even regno.
3470         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
3471         (aarch64_atomic_ldop_supported_p): Remove.
3472         (aarch64_gen_atomic_ldop): Remove.
3473         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
3474         Fully expand LSE operations here.
3475         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
3476         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
3477         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
3478         and use ATOMIC_LDOP instead; use register_operand for the input;
3479         drop the split and emit insns directly.
3480         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
3481         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
3482         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
3484         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
3485         (aarch64_gen_atomic_ldop): Don't call it.
3486         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
3487         Use aarch64_reg_or_zero.
3488         (aarch64_atomic_exchange<ALLI>): Likewise.
3489         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
3490         operand 0; use aarch64_reg_or_zero for input; merge ...
3491         (@aarch64_atomic_swp<ALLI>): ... this and remove.
3493         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
3494         (aarch64_split_compare_and_swap): Use it.
3495         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
3496         test oldval against the proper predicate.
3497         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
3498         Use nonmemory_operand for expected.
3499         (cas_short_expected_pred): New.
3500         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
3501         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
3502         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
3503         (aarch64_plushi_operand): New.
3505         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
3506         Force oldval into the rval register for TARGET_LSE; emit the compare
3507         during initial expansion so that it may be deleted if unused.
3508         (aarch64_gen_atomic_cas): Remove.
3509         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
3510         Change =&r to +r for operand 0; use match_dup for operand 2;
3511         remove is_weak and mod_f operands as unused.  Drop the split
3512         and merge with...
3513         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
3514         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
3515         (@aarch64_atomic_cas<GPI>): Similarly.
3517 2018-10-31  Richard Biener  <rguenther@suse.de>
3519         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
3520         using ABSU_EXPR.
3522 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3524         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
3525         * config/aarch64/aarch64.md: Include saphira.md
3526         * config/aarch64/saphira.md: New file for pipeline description.
3528 2018-10-30  Martin Sebor  <msebor@redhat.com>
3530         PR middle-end/87041
3531         * gimple-ssa-sprintf.c (format_directive): Use %G to include
3532         inlining context.
3533         (sprintf_dom_walker::compute_format_length):
3534         Avoid setting POSUNDER4K here.
3535         (get_destination_size): Handle null argument values.
3536         (get_user_idx_format): New function.
3537         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
3538         functions, including user-defined with attribute format printf.
3539         Use %G to include inlining context.
3540         Set POSUNDER4K here.
3542 2018-10-30  Jan Hubicka  <jh@suse.cz>
3544         * params.def (lto-partitions): Bump from 32 to 128.
3546 2018-10-30  Jan Hubicka  <jh@suse.cz>
3548         * tree.c
3549         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
3550         head in file.
3551         (free_lang_data_in_type): Forward declare.
3552         (fld_type_variant_equal_p): New function.
3553         (fld_type_variant): New function
3554         (fld_incomplete_types): New hash.
3555         (fld_incomplete_type_of): New function
3556         (fld_simplfied-type): New function.
3557         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
3558         (free_lang_data): Allocate and free fld_incomplete_type; update call
3559         of free_lang_data_in_decl.
3561 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
3563         * gcov.c (output_lines): Remove duplicate line.
3565 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
3567         * config/rs6000/rs6000.md (bswapdi2): Force address into register
3568         if not in indexed or indirect form.
3569         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
3570         (bswapdi2_store): Ditto.
3571         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
3572         helper function.
3573         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
3574         Prototype for helper function.
3576 2018-10-30  Martin Sebor  <msebor@redhat.com>
3578         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
3579         (target, pragma GCC optimize, pragma GCC target): Ditto.
3581 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
3583         * doc/extend.texi: Fix prototype and description of
3584         __builtin_expect_with_probability.
3586 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
3588         * cgraph.h (clone_function_name_1): Replaced by new
3589           clone_function_name_numbered that takes name as string; for
3590           privatize_symbol_name_1 use only.
3591           (clone_function_name): Renamed to
3592           clone_function_name_numbered to be explicit about numbering.
3593           (clone_function_name): New two-argument function that does
3594           not number its output.
3595           (clone_function_name): New three-argument function that
3596           takes a number to append to its output.
3597         * cgraphclones.c (duplicate_thunk_for_node):
3598           (clone_function_name_1): Renamed.
3599           (clone_function_name_numbered): Two new functions.
3600           (clone_function_name): Improved documentation.
3601           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
3602         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
3603         * final.c (final_scan_insn_1): Use the new clone_function_name
3604           without numbering.
3605         * multiple_target.c (create_dispatcher_calls): Ditto.
3606           (create_target_clone): Ditto.
3607         * omp-expand.c (grid_expand_target_grid_body): Ditto.
3608         * omp-low.c (create_omp_child_function_name): Ditto.
3609         * omp-simd-clone.c (simd_clone_create): Ditto.
3610         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
3611           new clone_function_name without numbering.
3613 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
3615         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
3616         Assert that the allocation size is not zero.
3618 2018-10-30  Richard Biener  <rguenther@suse.de>
3620         PR tree-optimization/87800
3621         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
3622         non-induction or reduction PHIs.
3624 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
3626         * config/aarch64/falkor-tag-collision-avoidance.c
3627         (execute_tag_collision_avoidance): Call df_note_add_problem.
3629 2018-10-30  Martin Liska  <mliska@suse.cz>
3631         * doc/extend.texi: Fix typo in documentation
3632         of __builtin_expect_with_probability.
3634 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
3636         PR c++/87721
3637         * input.c (get_substring_ranges_for_loc): Detect if
3638         linemap_resolve_location gives us a NULL map, and reject
3639         this case.
3641 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
3643         * config.gcc (xstormy16-*-elf): Set tm_d_file.
3645 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
3646             Martin Sebor  <msebor@redhat.com>
3647             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3649         * doc/ux.texi (Quoting): New subsection, adapted from material at
3650         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
3651         MartinSebor and ManuelLopezIbanez.
3652         (Fix-it hints): Note that fix-it hints shouldn't be marked for
3653         translation.
3655 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
3657         PR middle-end/87469
3658         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
3659         max value.
3661 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3663         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
3665 2018-10-29  Paul Koning  <ni1d@arrl.net>
3667         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
3669 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3671         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
3672         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
3673         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
3674         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
3675         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
3676         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
3677         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
3678         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
3679         Change 'vector' to '__vector'.
3680         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
3681         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
3682         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
3683         _mm_avg_pu16): Likewise.  And, whitespace corrections.
3685 2018-10-29  Richard Biener  <rguenther@suse.de>
3687         PR tree-optimization/87785
3688         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3689         internal defs.
3691 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3693         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3695 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3697         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3698         ports configuration macro, defaults to "gnu".
3699         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3700         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3702 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3704         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3706 2018-10-29  Richard Biener  <rguenther@suse.de>
3708         PR tree-optimization/87790
3709         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3710         (vect_make_slp_decision): Adjust.
3711         (vect_slp_analyze_bb_1): Likewise.
3712         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3713         edges.
3715 2018-10-29  Richard Biener  <rguenther@suse.de>
3717         PR tree-optimization/87785
3718         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3719         and processing.
3720         (vect_build_slp_tree): Likewise.
3721         (vect_gather_slp_loads): New function.
3722         (vect_analyze_slp_instance): Gather loads separately from the
3723         SLP tree build.
3725 2018-10-29  Martin Liska  <mliska@suse.cz>
3727         * Makefile.in: Make dependency to json.o.
3728         * doc/gcov.texi: Document new JSON format, remove
3729         old intermediate format documentation.
3730         * gcov.c (struct function_info): Come up with m_name and
3731         m_demangled_name.
3732         (function_info::function_info): Initialize it.
3733         (function_info::~function_info): Release it.
3734         (main): Rename flag_intermediate_format to flag_json_format.
3735         (print_usage): Describe --json-format.
3736         (process_args): Set flag_json_format.
3737         (output_intermediate_line): Remove.
3738         (output_intermediate_json_line): Likewise.
3739         (get_gcov_intermediate_filename): Return new extension
3740         ".gcov.json.gz".
3741         (output_intermediate_file): Implement JSON emission.
3742         (output_json_intermediate_file): Implement JSON emission.
3743         (generate_results): Use ::get_name for function name.
3744         Handle JSON output file.
3745         (read_graph_file): Use ::get_name instead of cplus_demangle.
3746         (read_count_file): Likewise.
3747         (solve_flow_graph): Likewise.
3748         (add_line_counts): Likewise.
3749         (accumulate_line_counts): Use new flag_json_format.
3750         (output_function_details): Use ::get_name instead of cplus_demangle.
3751         (output_lines): Likewise.
3752         * json.cc (test_writing_literals): Add new tests.
3753         * json.h (class literal): Add new boolean constructor.
3755 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
3757         PR rtl-optimization/87701
3758         PR rtl-optimization/87780
3759         * combine.c (make_more_copies): Rewrite.
3761 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
3763         * doc/generic.texi (ABSU_EXPR): Document.
3764         * match.pd (absu(x)*absu(x) -> x*x): Handle.
3765         (absu(absu(X)) -> absu(X)): Likewise.
3766         (absu(-X) -> absu(X)): Likewise.
3767         (absu(X)  where X is nonnegative -> X): Likewise.
3769 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
3771         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3772         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3773         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3774         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3775         (s-tm-texi): Also check timestamp on d-target.def.
3776         (generated_files): Add TM_D_H and d-target-hooks-def.h.
3777         (build/genhooks.o): Also depend on D_TARGET_DEF.
3778         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3779         variables.
3780         * config/aarch64/aarch64-d.c: New file.
3781         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3782         Define.
3783         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3784         prototype.
3785         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3786         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3787         * config/arm/arm-d.c: New file.
3788         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3789         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3790         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3791         * config/arm/t-arm (arm-d.o): New rule.
3792         * config/default-d.c: New file.
3793         * config/glibc-d.c: New file.
3794         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3795         * config/i386/i386-d.c: New file.
3796         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3797         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3798         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3799         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3800         * config/i386/t-i386 (i386-d.o): New rule.
3801         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3802         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3803         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3804         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3805         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3806         * config/mips/mips-d.c: New file.
3807         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3808         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3809         * config/mips/t-mips (mips-d.o): New rule.
3810         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3811         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3812         * config/powerpcspe/powerpcspe-d.c: New file.
3813         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3814         New prototype.
3815         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3816         Support GNU D by using 0 as the language type.
3817         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3818         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3819         * config/riscv/riscv-d.c: New file.
3820         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3821         prototype.
3822         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3823         * config/riscv/t-riscv (riscv-d.o): New rule.
3824         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3825         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3826         * config/rs6000/rs6000-d.c: New file.
3827         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3828         prototype.
3829         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3830         Support GNU D by using 0 as the language type.
3831         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3832         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3833         * config/s390/s390-d.c: New file.
3834         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3835         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3836         * config/s390/t-s390 (s390-d.o): New rule.
3837         * config/sparc/sparc-d.c: New file.
3838         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3839         prototype.
3840         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3841         * config/sparc/t-sparc (sparc-d.o): New rule.
3842         * config/t-glibc (glibc-d.o): New rule.
3843         * configure: Regenerated.
3844         * configure.ac (tm_d_file): New variable.
3845         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3846         * doc/contrib.texi (Contributors): Add self for the D frontend.
3847         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3848         * doc/install.texi (Configuration): Mention libphobos as an option for
3849         --enable-shared.  Mention d as an option for --enable-languages.
3850         (Testing): Mention check-d as a target.
3851         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3852         name suffixes.  Mention d as a -x option.
3853         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3854         * doc/standards.texi (Standards): Add section on D language.
3855         * doc/tm.texi: Regenerated.
3856         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3857         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3858         * dwarf2out.c (is_dlang): New function.
3859         (gen_compile_unit_die): Use DW_LANG_D for D.
3860         (declare_in_namespace): Return module die for D, instead of adding
3861         extra declarations into the namespace.
3862         (gen_namespace_die): Generate DW_TAG_module for D.
3863         (gen_decl_die): Handle CONST_DECLSs for D.
3864         (dwarf2out_decl): Likewise.
3865         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3866         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3867         kinds of aggregates.
3868         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3869         * genhooks.c: Include d/d-target.def.
3871 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3873         PR target/85669
3874         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3875         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3876         (STACK_DYNAMIC_OFFSET): Likewise.
3878 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3880         PR target/80024
3881         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3882         error message.
3884 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3886         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3887         return type and other typos.
3889 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3891         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3892         a shorter sequence with fewer branches.
3893         (emit_final_str_compare_gpr): Ditto.
3895 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3897         * config/rs6000/tmmintrin.h: New file.
3898         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3900 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3902         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3903         * config/rs6000/xmmintrin.h: Likewise.
3905 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3907         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3909 2018-10-26  Richard Biener  <rguenther@suse.de>
3911         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3912         and wrapper.
3913         (vect_mark_slp_stmts_relevant): Likewise.
3914         (vect_detect_hybrid_slp_stmts): Likewise.
3915         (vect_bb_slp_scalar_cost): Likewise.
3916         (vect_remove_slp_scalar_calls): Likewise.
3918 2018-10-26  Jan Hubicka  <jh@suse.cz>
3920         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3921         (warn_types_mismatch): Fix walk of DECL_NAME.
3922         (odr_types_equivalent_p): Fix overactive assert.
3924 2018-10-26  Richard Biener  <rguenther@suse.de>
3926         PR tree-optimization/87105
3927         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3928         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3929         refcnt.
3930         (vect_create_new_slp_node): Initialize refcnt to one.
3931         (bst_traits): Move.
3932         (scalar_stmts_set_t, bst_fail): Remove.
3933         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3934         (vect_build_slp_tree): Add bst_map argument and lookup
3935         already created SLP nodes.
3936         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3937         addresses.
3938         (vect_slp_rearrange_stmts): Handle a SLP graph.
3939         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3940         the CSE map.  Fix indenting.
3941         (vect_schedule_slp_instance): Add short-cut.
3943 2018-10-26  Martin Liska  <mliska@suse.cz>
3945         PR testsuite/86158
3946         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3947         addr_expr and not with pointers.
3949 2018-10-26  Jan Hubicka  <jh@suse.cz>
3951         * tree.c (free_lang_data_in_type): Only check main variants.
3952         * ipa-devirt.c (warn_odr): Make static.
3953         (types_same_for_odr): Drop strict variant.
3954         (types_odr_comparable): Likewise.
3955         (odr_or_derived_type_p): Look for main variants.
3956         (odr_name_hasher::equal): Cleanup comment.
3957         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3958         variants.
3959         (type_variants_equivalent_p): break out from ...
3960         (odr_types_equivalent): ... here; go for main variants where needed.
3961         (warn_odr): ... here; turn static.
3962         (warn_types_mismatch): Compare mangled names of main variants.
3963         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3964         (type_with_linkage_p): Sanity check that we look at main variant.
3965         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3966         * tree.h (types_same_for_odr): Drop strict argument.
3968 2018-10-26  Richard Biener  <rguenther@suse.de>
3970         PR tree-optimization/87746
3971         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3972         Simplify and fix WRT strided store groups with size not
3973         equal to step in element count.
3974         (vect_analyze_group_access_1): Dump the whole group.
3976 2018-10-25  Carl Love  <cel@us.ibm.com>
3978         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3979         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3980         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3981         precicion entry for each overloaded builtin.
3982         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3983         VSCEDPUO): Rename overloaded name.
3984         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3985         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3986         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3987         define_expand for xscmpexqp instruction.
3988         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3990 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3991             Jinsong Ji <jji@us.ibm.com>
3993         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3994         function with vec_sl.
3995         (_mm_slli_epi32): Likewise.
3996         (_mm_slli_epi64): Likewise.
3997         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3998         (_mm_srai_epi32): Likewise.
3999         (_mm_srli_epi16): Replace deprecated function with vec_sr.
4000         (_mm_srli_epi32): Likewise.
4001         (_mm_srli_epi64): Likewise.
4002         (_mm_sll_epi16): Replace deprecated function with vec_sl.
4003         (_mm_sll_epi32): Likewise.
4004         (_mm_sll_epi64): Likewise.
4005         (_mm_sra_epi16): Replace deprecated function with vec_sra.
4006         (_mm_sra_epi32): Likewise.
4007         (_mm_srl_epi16): Replace deprecated function with vec_sr.
4008         (_mm_srl_epi32): Likewise.
4009         (_mm_srl_epi64): Likewise.
4011 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4012             Jinsong Ji <jji@us.ibm.com>
4014         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
4015         comparison operators with vec_cmp* for compatibility due to
4016         unfortunate history; clean up formatting and use types more
4017         appropriately.
4018         (_mm_sll_epi32): Likewise.
4019         (_mm_sll_epi64): Likewise.
4020         (_mm_srl_epi16): Likewise.
4021         (_mm_srl_epi32): Likewise.
4022         (_mm_srl_epi64): Likewise.
4024 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4025             Jinsong Ji <jji@us.ibm.com>
4027         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
4028         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
4029         __vector __bool int.  Use vec_cmpgt in preference to deprecated
4030         function vec_vcmpgtfp.
4031         (_mm_max_ps): Likewise.
4033 2018-10-25  Jeff Law  <law@redhat.com>
4035         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
4036         if returning false.
4038 2018-10-25  Martin Sebor  <msebor@redhat.com>
4040         * doc/extend.texi (aligned): Expand attribute description.
4041         (Alignment): Rename section.  Discuss function arguments.
4043 2018-10-25  Jan Hubicka  <jh@suse.cz>
4045         * ipa-devirt.c (main_odr_variant): Remove.
4046         (hash_odr_name, types_same_for_odr, types_odr_comparable,
4047         odr_name_hasher::equal, odr_subtypes_equivalent_p):
4048         Drop use of main_odr_variant.
4049         (add_type_duplicate): Silence confused warnings on integer types.
4050         (get_odr_type): Always look for main variant.
4051         (register_odr_type): Simplify.
4053 2018-10-25  Richard Biener  <rguenther@suse.de>
4055         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
4056         Initialize ng to silence error with release checking bootstrap.
4058 2018-10-25  Richard Biener  <rguenther@suse.de>
4060         * tree-if-conv.c: Include tree-ssa-sccvn.h.
4061         (tree_if_conversion): Run CSE on the if-converted loop body.
4063 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4065         * config/s390/constraints.md (ZL): New constraint.
4066         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
4067         operands.
4068         * config/s390/s390.md (movdi_larl): Remove.
4069         (movdi_64): Add the LARL alternative.
4071 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4073         PR bootstrap/87747
4074         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
4075         (RTX_CODE_HWINT_P): New macro.
4076         (rtx_code_size): Use RTX_CODE_HWINT_P ().
4078 2018-10-25  Jan Hubicka  <jh@suse.cz>
4080         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
4081         is anonymous.
4083 2018-10-25  Richard Biener  <rguenther@suse.de>
4085         PR tree-optimization/87665
4086         PR tree-optimization/87745
4087         * tree-vectorizer.h (get_earlier_stmt): Remove.
4088         (get_later_stmt): Pick up UID from the original non-pattern stmt.
4090 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
4092         * options.texi (Deprecated): Move list to Var section.
4094 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
4095             Jinsong Ji <jji@us.ibm.com>
4097         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
4098         __vector long to __vector long long.
4099         (_mm_cvtpd_ps): Likewise.
4100         (_mm_cvttpd_epi32): Likewise.
4101         (_mm_cvtpi32_pd): Likewise.
4102         (_mm_unpackhi_epi64): Likewise.
4103         (_mm_unpacklo_epi64): Likewise.
4105 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
4107         PR rtl-optimization/87720
4108         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
4110 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
4112         * gimple-ssa-isolate-paths.c
4113         (find_implicit_erroneous_behavior): Do not change code if the
4114         pass is running for warnings only.
4115         (find_explicit_erroneous_behavior): Likewise.
4117 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
4119         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
4120         Define as rs6000_mangle_decl_assembler_name.
4121         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
4122         long double to IEEE long double, switch the names of the long
4123         double built-in functions to be <func>f128 instead of <func>l.
4125 2018-10-24  Martin Sebor  <msebor@redhat.com>
4127         * doc/extend.texi (nonnull): List no-argument form.  Reference
4128         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
4130 2018-10-24  Richard Biener  <rguenther@suse.de>
4132         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
4134 2018-10-24  Martin Liska  <mliska@suse.cz>
4136         PR tree-optimization/84436
4137         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
4138         Remove.
4139         (switch_conversion::contains_linear_function_p): New.
4140         (switch_conversion::build_one_array): Support linear
4141         transformation on input.
4142         * tree-switch-conversion.h (struct switch_conversion): Add
4143         contains_linear_function_p declaration.
4145 2018-10-24  Richard Biener  <rguenther@suse.de>
4147         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
4148         if its argument is CONSTANT_CLASS_P.
4150 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4152         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
4153         it is wrong for forward declarations.
4155 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4157         * config/s390/s390.c (s390_check_qrst_address): Add the missing
4158         SYMBOL_REF_P () check.
4160 2018-10-24  Richard Biener  <rguenther@suse.de>
4162         PR tree-optimization/87105
4163         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
4164         dump classification.
4165         (vect_analyze_data_ref_accesses): Handle duplicate loads and
4166         stores by splitting the affected group after the fact.
4167         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
4168         fail the SLP build because of size constraints.
4170 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4172         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
4173         * configure: Regenerate.
4174         * config.in: Regenerate.
4175         * varasm.c (mergeable_string_section): Use readonly_data_section
4176         if linker doesn't support SHF_MERGE with alignment > 8.
4177         (mergeable_constant_section): Likewise.
4179 2018-10-24  Richard Biener  <rguenther@suse.de>
4181         PR tree-optimization/84013
4182         * tree-ssa-structalias.c (struct msdi_data): New struct for
4183         marshalling data to walk_stmt_load_store_ops.
4184         (maybe_set_dependence_info): Refactor as callback for
4185         walk_stmt_load_store_ops.
4186         (compute_dependence_clique): Set restrict info on all stmt kinds.
4188 2018-10-24  Martin Liska  <mliska@suse.cz>
4190         * cgraph.c (cgraph_node::dump):
4191         Remove reduntant dumps and make tp_first_run dump more compact.
4193 2018-10-24  Richard Biener  <rguenther@suse.de>
4195         PR tree-optimization/87665
4196         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
4197         to reflect reality.
4199 2018-10-12  Jeff Law  <law@redhat.com>
4201         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
4202         for H8/S.
4204 2018-10-23  Richard Biener  <rguenther@suse.de>
4206         * tree-vrp.c (add_assert_info): Guard dump_printf with
4207         dump_enabled_p.
4208         * gimple-ssa-evrp-analyze.c
4209         (evrp_range_analyzer::record_ranges_from_incoming_edge):
4210         Use value_range::ignore_equivs_equal_p.
4212 2018-10-23  Richard Biener  <rguenther@suse.de>
4214         PR tree-optimization/87105
4215         PR tree-optimization/87608
4216         * passes.def (pass_all_early_optimizations): Add early phi-opt
4217         after dce.
4218         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
4219         addition to debug stmts.
4220         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
4221         and abs replacement early.
4222         * tree-cfg.c (gimple_empty_block_p): Likewise.
4224 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
4226         PR target/86383
4227         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
4228         specified to configure.
4229         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
4231 2018-10-23  Richard Biener  <rguenther@suse.de>
4233         PR tree-optimization/87700
4234         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
4236 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
4238         PR target/87674
4239         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
4240         second argument from __mmask16 to __mmask8.
4241         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
4242         _mm_mask_packs_epi32): Likewise.
4243         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
4244         Likewise.
4245         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
4247 2018-10-23  Richard Biener  <rguenther@suse.de>
4249         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
4251 2018-10-23  Richard Biener  <rguenther@suse.de>
4253         PR tree-optimization/86144
4254         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
4255         over simd attribute.
4257 2018-10-23  Richard Biener  <rguenther@suse.de>
4259         PR tree-optimization/87693
4260         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
4261         the case we do not find the taken edge.
4263 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
4264             Jinsong Ji  <jji@us.ibm.com>
4266         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
4267         (_mm_store_pd): Use unaligned vector type for pointer cast.
4268         (_mm_maskmoveu_si128): Likewise.
4269         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
4270         (_mm_store_ps): Use unaligned vector type for pointer cast.
4272 2018-10-22  Paul Koning  <ni1d@arrl.net>
4274         * symtab.c (symtab_node::increase_alignment): Correct max
4275         alignment check.
4277 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
4279         PR tree-optimization/87633
4280         * match.pd: Do not generate unordered integer comparisons.
4282 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
4284         PR rtl-optimization/87600
4285         * combine.c: Add include of expr.h.
4286         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
4287         register to a pseudo.
4288         (make_more_copies): New function, add a copy to a new pseudo after
4289         the moves from hard registers into pseudos.
4290         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
4291         later.  Call make_more_copies.
4293 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
4295         * lra-constraints.c (process_alt_operands): New local array,
4296         matching_early_clobber.  Check matching_early_clobber before
4297         decrementing reject, and set matching_early_clobber after.
4299 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
4301         PR target/87598
4302         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
4303         call output_addr_const and hope for the best.
4305 2018-10-22  Richard Biener  <rguenther@suse.de>
4307         * gimple-ssa-evrp-analyze.c
4308         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
4309         smarter about what ranges to use.
4310         * tree-vrp.c (add_assert_info): Dump here.
4311         (register_edge_assert_for_2): Instead of here at multiple but
4312         not all places.
4314         * gcc.dg/tree-ssa/evrp12.c: New testcase.
4315         * gcc.dg/predict-6.c: Adjust.
4316         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
4317         * gcc.dg/tree-ssa/vrp02.c: Likewise.
4318         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
4320 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
4321         Richard Biener  <rguenther@suse.de>
4323         * bitmap.h: Update data structure documentation, including a
4324         description of bitmap views as either linked-lists or splay trees.
4325         (struct bitmap_element_def): Update comments for splay tree bitmaps.
4326         (struct bitmap_head_def): Likewise.
4327         (bitmap_list_view, bitmap_tree_view): New prototypes.
4328         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
4329         tree_form fields.
4330         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
4331         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
4332         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
4333         released bitmap element here.
4334         (bitmap_element_free): Remove.
4335         (bitmap_elt_clear_from): Work on splay tree bitmaps.
4336         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
4337         this function similar ones such that linked-list bitmap implementation
4338         functions are grouped.
4339         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
4340         and moved for grouping.
4341         (bitmap_list_insert_element_after): Renamed from
4342         bitmap_elt_insert_after, and moved for grouping.
4343         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
4344         (bitmap_tree_link_left, bitmap_tree_link_right,
4345         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
4346         bitmap_tree_link_element, bitmap_tree_unlink_element,
4347         bitmap_tree_find_element): New functions for splay-tree bitmap
4348         implementation.
4349         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
4350         Renamed and moved, see above entries.
4351         (bitmap_tree_listify_from): New function to convert part of a splay
4352         tree bitmap to a linked-list bitmap.
4353         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
4354         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
4355         (bitmap_find_bit): Remove.
4356         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
4357         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
4358         Handle splay tree bitmaps.
4359         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
4360         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
4361         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
4362         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
4363         bitmap_intersect_compl_p, bitmap_ior_and_compl,
4364         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
4365         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
4366         corresponding changes to use linked-list specific bitmap_element
4367         manipulation functions as applicable for efficiency.
4368         (bitmap_tree_to_vec): New function.
4369         (debug_bitmap_elt_file): New function split out from ...
4370         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
4371         (bitmap_print): Likewise.
4373         PR tree-optimization/63155
4374         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
4375         SSA edge worklists.
4376         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
4377         in tree-view.
4379 2018-10-22  Martin Liska  <mliska@suse.cz>
4381         PR tree-optimization/87686
4382         Revert
4383         2018-08-29  Martin Liska  <mliska@suse.cz>
4385         * tree-switch-conversion.c (switch_conversion::expand):
4386         Strenghten assumption about gswitch statements.
4388 2018-10-22  Martin Liska  <mliska@suse.cz>
4390         * ipa-icf.c (sem_item::compare_attributes): Remove.
4391         (sem_item::compare_referenced_symbol_properties): Use
4392         attribute_list_equal instead.
4393         (sem_function::equals_wpa): Likewise.
4394         * ipa-icf.h: Remove compare_attributes.
4396 2018-10-22  Richard Biener  <rguenther@suse.de>
4398         PR middle-end/87682
4399         * mem-stats.h (mem_usage::operator==): Fix pasto.
4401 2018-10-22  Richard Biener  <rguenther@suse.de>
4403         PR tree-optimization/87640
4404         * tree-vrp.c (set_value_range_with_overflow): Decompose
4405         incomplete result.
4406         (extract_range_from_binary_expr_1): Adjust.
4408 2018-10-22  Martin Jambor  <mjambor@suse.cz>
4410         * tree-eh.h (stmt_could_throw_p): Add function parameter.
4411         (stmt_can_throw_external): Likewise.
4412         (stmt_can_throw_internal): Likewise.
4413         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
4414         (lower_eh_constructs_2): Likewise.
4415         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
4416         (stmt_can_throw_external): Likewise.
4417         (stmt_can_throw_internal): Likewise.
4418         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
4419         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
4420         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
4421         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
4422         (pass_lower_eh_dispatch::execute): Pass cfun to
4423         stmt_can_throw_external.
4424         (cleanup_empty_eh): Likewise.
4425         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
4426         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
4427         stmt_can_throw_external instead of pushing it to cfun.
4428         (symbol_table::create_edge): Likewise.
4429         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
4430         stmt_can_throw_internal.
4431         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
4432         to stmt_could_throw_p.
4433         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
4434         stmt_can_throw_internal.
4435         (pass_store_merging::execute): Likewise.
4436         * gimple-ssa-strength-reduction.c
4437         (find_candidates_dom_walker::before_dom_children): Pass cfun to
4438         stmt_could_throw_p.
4439         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
4440         stmt_can_throw_internal.
4441         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
4442         to stmt_can_throw_external.
4443         (check_stmt): Pass cfun to stmt_could_throw_p.
4444         (check_stmt): Pass cfun to stmt_can_throw_external.
4445         (pass_nothrow::execute): Likewise.
4446         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
4447         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
4448         stmt_can_throw_internal.
4449         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
4450         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
4451         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
4452         * tree-complex.c (expand_complex_libcall): Pass cfun to
4453         stmt_could_throw_p and to stmt_can_throw_internal.
4454         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
4455         * tree-inline.c (copy_edges_for_bb): Likewise.
4456         (maybe_move_debug_stmts_to_successors): Likewise.
4457         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
4458         stmt_could_throw_p.
4459         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
4460         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
4461         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
4462         stmt_can_throw_internal.
4463         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
4464         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
4465         stmt_could_throw_p.
4466         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
4467         stmt_can_throw_internal.
4468         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
4469         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
4470         stmt_could_throw_p.
4471         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
4472         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
4473         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4474         (convert_mult_to_fma_1): Likewise.
4475         (convert_to_divmod): Likewise.
4476         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
4477         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
4478         * tree-ssa-propagate.c
4479         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
4480         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
4481         (maybe_optimize_range_tests): Likewise.
4482         (linearize_expr_tree): Likewise.
4483         (reassociate_bb): Likewise.
4484         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4485         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
4486         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
4487         (handle_char_store): Likewise.
4488         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
4489         stmt_can_throw_internal.
4490         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
4491         stmt_could_throw_p.
4492         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4493         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
4494         (vectorizable_simd_clone_call): Likewise.
4495         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
4496         (gimple_stringop_fixed_value): Likewise.
4498 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
4500         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
4501         literal pool references.
4502         (s390_check_qrst_address): Adapt to the new behavior of
4503         s390_loadrelative_operand_p ().
4505 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
4507         PR target/72782
4508         * config/i386/sse.md (*andnot<mode>3_bcst): New.
4510 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
4512         PR target/72782
4513         * config/i386/sse.md (*<code><mode>3_bcst): New.
4515 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
4517         PR target/72782
4518         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
4519         V4DI, V16SI and V8DI.
4520         (*sub<mode>3<mask_name>_bcst): New.
4521         (*add<mode>3<mask_name>_bcst): Likewise.
4523 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
4524             Jinsong Ji  <jji@us.ibm.com>
4526         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
4527         __m64 with __vector unsigned long long for compatibility.
4528         (_mm_movemask_epi8): Likewise.
4529         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
4530         (_mm_cvttps_pi32): Likewise.
4531         (_mm_cvtpi32_ps): Likewise.
4532         (_mm_cvtps_pi16): Likewise.
4533         (_mm_loadh_pi): Likewise.
4534         (_mm_storeh_pi): Likewise.
4535         (_mm_movehl_ps): Likewise.
4536         (_mm_movelh_ps): Likewise.
4537         (_mm_loadl_pi): Likewise.
4538         (_mm_storel_pi): Likewise.
4539         (_mm_movemask_ps): Likewise.
4540         (_mm_shuffle_pi16): Likewise.
4542 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4544         PR target/72782
4545         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
4546         __builtin_ia32_vfnmsubpd512_mask.
4547         (_mm512_mask_fnmsub_round_pd): Likewise.
4548         (_mm512_fnmsub_pd): Likewise.
4549         (_mm512_mask_fnmsub_pd): Likewise.
4550         (_mm512_maskz_fnmsub_round_pd): Use
4551         __builtin_ia32_vfnmsubpd512_maskz.
4552         (_mm512_maskz_fnmsub_pd): Likewise.
4553         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
4554         (_mm512_mask_fnmsub_round_ps): Likewise.
4555         (_mm512_fnmsub_ps): Likewise.
4556         (_mm512_mask_fnmsub_ps): Likewise.
4557         (_mm512_maskz_fnmsub_round_ps): Use
4558         __builtin_ia32_vfnmsubps512_maskz.
4559         (_mm512_maskz_fnmsub_ps): Likewise.
4560         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
4561         __builtin_ia32_vfnmsubpd256_mask.
4562         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
4563         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4564         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
4565         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
4566         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
4567         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
4568         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
4569         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
4570         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
4571         __builtin_ia32_vfnmsubpd.
4572         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
4573         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
4574         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
4575         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
4576         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
4577         * config/i386/i386-builtin.def: Add
4578         __builtin_ia32_vfnmsubpd256_mask,
4579         __builtin_ia32_vfnmsubpd256_maskz,
4580         __builtin_ia32_vfnmsubpd128_mask,
4581         __builtin_ia32_vfnmsubpd128_maskz,
4582         __builtin_ia32_vfnmsubps256_mask,
4583         __builtin_ia32_vfnmsubps256_maskz,
4584         __builtin_ia32_vfnmsubps128_mask,
4585         __builtin_ia32_vfnmsubps128_maskz,
4586         __builtin_ia32_vfnmsubpd512_mask,
4587         __builtin_ia32_vfnmsubpd512_maskz,
4588         __builtin_ia32_vfnmsubps512_mask,
4589         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
4590         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
4591         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
4592         __builtin_ia32_vfnmsubpd256.
4593         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
4594         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
4595         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
4596         Likewise.
4597         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
4598         Likewise.
4599         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
4600         Likewise.
4601         (fmai_vmfnmsub_<mode><round_name>): Likewise.
4603 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4605         PR target/72782
4606         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
4607         __builtin_ia32_vfnmaddpd512_mask.
4608         (_mm512_mask_fnmadd_round_pd): Likewise.
4609         (_mm512_fnmadd_pd): Likewise.
4610         (_mm512_mask_fnmadd_pd): Likewise.
4611         (_mm512_maskz_fnmadd_round_pd): Use
4612         __builtin_ia32_vfnmaddpd512_maskz.
4613         (_mm512_maskz_fnmadd_pd): Likewise.
4614         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
4615         (_mm512_mask_fnmadd_round_ps): Likewise.
4616         (_mm512_fnmadd_ps): Likewise.
4617         (_mm512_mask_fnmadd_ps): Likewise.
4618         (_mm512_maskz_fnmadd_round_ps): Use
4619         __builtin_ia32_vfnmaddps512_maskz.
4620         (_mm512_maskz_fnmadd_ps): Likewise.
4621         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
4622         __builtin_ia32_vfnmaddpd256_mask.
4623         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
4624         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
4625         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
4626         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4627         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4628         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
4629         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
4630         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
4631         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
4632         __builtin_ia32_vfnmaddpd.
4633         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
4634         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
4635         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
4636         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
4637         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
4638         * config/i386/i386-builtin.def: Add
4639         __builtin_ia32_vfnmaddpd256_mask,
4640         __builtin_ia32_vfnmaddpd256_maskz,
4641         __builtin_ia32_vfnmaddpd128_mask,
4642         __builtin_ia32_vfnmaddpd128_maskz,
4643         __builtin_ia32_vfnmaddps256_mask,
4644         __builtin_ia32_vfnmaddps256_maskz,
4645         __builtin_ia32_vfnmaddps128_mask,
4646         __builtin_ia32_vfnmaddps128_maskz,
4647         __builtin_ia32_vfnmaddpd512_mask,
4648         __builtin_ia32_vfnmaddpd512_maskz,
4649         __builtin_ia32_vfnmaddps512_mask,
4650         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
4651         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
4652         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
4653         __builtin_ia32_vfnmaddpd256.
4654         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
4655         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
4656         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
4657         Likewise.
4658         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
4659         Likewise.
4660         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
4661         Likewise.
4662         (fmai_vmfnmadd_<mode><round_name>): Likewise.
4664 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4666         PR target/72782
4667         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
4668         __builtin_ia32_vfmsubpd512_mask.
4669         (_mm512_mask_fmsub_round_pd): Likewise.
4670         (_mm512_fmsub_pd): Likewise.
4671         (_mm512_mask_fmsub_pd): Likewise.
4672         (_mm512_maskz_fmsub_round_pd): Use
4673         __builtin_ia32_vfmsubpd512_maskz.
4674         (_mm512_maskz_fmsub_pd): Likewise.
4675         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
4676         (_mm512_mask_fmsub_round_ps): Likewise.
4677         (_mm512_fmsub_ps): Likewise.
4678         (_mm512_mask_fmsub_ps): Likewise.
4679         (_mm512_maskz_fmsub_round_ps): Use
4680         __builtin_ia32_vfmsubps512_maskz.
4681         (_mm512_maskz_fmsub_ps): Likewise.
4682         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
4683         __builtin_ia32_vfmsubpd256_mask.
4684         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4685         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4686         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4687         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4688         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4689         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4690         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4691         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4692         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4693         __builtin_ia32_vfmsubpd.
4694         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4695         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4696         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4697         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4698         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4699         * config/i386/i386-builtin.def: Add
4700         __builtin_ia32_vfmsubpd256_mask,
4701         __builtin_ia32_vfmsubpd256_maskz,
4702         __builtin_ia32_vfmsubpd128_mask,
4703         __builtin_ia32_vfmsubpd128_maskz,
4704         __builtin_ia32_vfmsubps256_mask,
4705         __builtin_ia32_vfmsubps256_maskz,
4706         __builtin_ia32_vfmsubps128_mask,
4707         __builtin_ia32_vfmsubps128_maskz,
4708         __builtin_ia32_vfmsubpd512_mask,
4709         __builtin_ia32_vfmsubpd512_maskz,
4710         __builtin_ia32_vfmsubps512_mask,
4711         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4712         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4713         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4714         __builtin_ia32_vfmsubpd256.
4715         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4716         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4717         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4718         Likewise.
4719         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4720         Likewise.
4721         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4722         Likewise.
4723         (fmai_vmfmsub_<mode><round_name>): Likewise.
4725 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4727         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4728         Remove plus.  Renamed to ...
4729         (*sub<mode>3<mask_name>_bcst): This.
4730         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4731         (*add<mode>3<mask_name>_bcst): This.
4733 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4735         PR target/72782
4736         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4738 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4740         PR target/87662
4741         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4742         (_mm_or_epi32): Likewise.
4743         (_mm256_xor_epi32): Likewise.
4744         (_mm_xor_epi32): Likewise.
4745         (_mm256_or_epi64): Likewise.
4746         (_mm_or_epi64): Likewise.
4747         (_mm256_xor_epi64): Likewise.
4748         (_mm_xor_epi64): Likewise.
4750 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
4752         PR target/72782
4753         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4755 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
4757         PR middle-end/87647
4758         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4760 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
4762         * doc/ux.texi: Move @section directly after @node.
4764 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
4766         PR middle-end/85488
4767         PR middle-end/87649
4768         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4769         depend closely nested inside of loop with ordered clause with
4770         a parameter.
4772 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
4774         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4775         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4776         * doc/ux.texi: New file.
4778 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
4780         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4781         be the first CR field allocated.
4783 2018-10-19  Richard Biener  <rguenther@suse.de>
4785         PR target/87657
4786         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4787         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4789 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4791         PR target/72782
4792         * config/i386/sse.md
4793         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4794         (*add<mode>3<mask_name>_bcst_2): Likewise.
4796 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4798         * config/i386/sse.md
4799         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4800         Replace nonimmediate_operand with register_operand.
4801         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4802         Likewise.
4803         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4804         Likewise.
4806 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4808         PR rtl-optimization/87596
4809         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4810         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4811         for instructions in FROM..TO range.
4813 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4815         * cfgexpand.c (expand_one_var): Use specific wording in error message
4816         for non-local frame variables.
4817         * stor-layout.c (layout_decl): Do not issue a warning for them.
4819 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4821         * haifa-sched.c (priority): Add force_recompute parameter.
4822         (apply_replacement): Call priority () with force_recompute = true.
4823         (restore_pattern): Likewise.
4825 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4827         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4828         HOST_BITS_PER_WIDE_INT.
4829         (test_vector_ops_duplicate): Likewise.
4831 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4833         PR target/72782
4834         * config/i386/sse.md (VF_AVX512): New.
4835         (avx512bcst): Likewise.
4836         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4837         Likewise.
4838         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4839         Likewise.
4840         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4841         Likewise.
4843 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4845         * doc/invoke.texi (-dumpversion): Improve grammar.
4846         (-dumpfullversion): Make more consistent with -dumpversion.
4848 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4850         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4851         Set JUMP_LABEL to the jump insn.
4852         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4853         Predict emitted jump and add label to jump insn.
4855 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4857         PR tree-optimization/87562
4858         * input.c (get_substring_ranges_for_loc): Use
4859         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4860         getting the linemap for the endpoint.  Verify that it's either
4861         in the same linemap as the start point's spelling location, or
4862         at least in the same file.
4864 2018-10-18  Richard Biener  <rguenther@suse.de>
4866         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4867         feed width-specific load/store costs through ix86_vec_cost.
4868         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4869         (k8_cost): Likewise.
4870         (bdver_cost): Likewise.
4871         (znver1_cost): Likewise.
4872         (btver1_cost): Likewise.
4873         (btver2_cost): Likewise.
4875 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4877         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4878         to simplify subreg of vec_merge.
4880 2018-10-18  Richard Biener  <rguenther@suse.de>
4882         * config/i386/i386.c: Fix costing of vector FMA.
4884 2018-10-18  Richard Biener  <rguenther@suse.de>
4886         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4887         and argument.
4888         (ix86_builtin_vectorization_cost): For vec_construct properly
4889         cost insertion into SSE regs.
4890         (...): Adjust calls to ix86_vec_cost.
4892 2018-10-18  Richard Biener  <rguenther@suse.de>
4894         PR middle-end/87087
4895         Revert
4896         2018-02-07  Richard Biener  <rguenther@suse.de>
4898         PR tree-optimization/84204
4899         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4900         this place.
4902 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4904         PR target/87537
4905         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4906         of vec_duplicate.
4907         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4908         VEC_MERGE of a VEC_DUPLICATE.
4910 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4912         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4913         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4914         * doc/standards.texi (C Language): Document C2X.
4915         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4916         (rl78_option_override): Handle "GNU C2X" language name.
4918 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4920         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4921         Document C17 as published in 2018.
4923 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4925         PR middle-end/87623
4926         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4927         bail out if both sides do not have the same storage order.
4929 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4931         * bitmap.c (bitmap_head::dump): New.
4932         * bitmap.h (bitmap_head): Add dump().
4933         * gimple-ssa-evrp-analyze.c
4934         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4935         (evrp_range_analyzer::set_ssa_range_info): Same.
4936         (evrp_range_analyzer::record_ranges_from_phis): Same.
4937         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4938         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4939         * gimple-ssa-sprintf.c (get_int_range): Same.
4940         (format_integer): Same.
4941         (sprintf_dom_walker::handle_gimple_call): Same.
4942         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4943         (ipcp_vr_lattice::top_p): Same.
4944         (ipcp_vr_lattice::bottom_p): Same.
4945         (ipcp_vr_lattice::set_to_bottom): Same.
4946         (ipa_vr_operation_and_type_effects): Same.
4947         (propagate_vr_across_jump_function): Same.
4948         (ipcp_store_vr_results): Same.
4949         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4950         (ipa_print_node_jump_functions_for_edge): Same.
4951         (ipa_get_value_range): Same.
4952         (ipa_compute_jump_functions_for_edge): Same.
4953         (ipa_write_jump_function): Same.
4954         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4955         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4956         Same.
4957         * vr-values.c (set_value_range_to_nonnegative): Same.
4958         (set_value_range_to_truthvalue): Same.
4959         (vr_values::get_value_range): Same.
4960         (vr_values::set_defs_to_varying): Same.
4961         (vr_values::update_value_range): Same.
4962         (symbolic_range_based_on_p): Same.
4963         (vr_values::op_with_boolean_value_range_p): Same.
4964         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4965         (vr_values::extract_range_from_ssa_name): Same.
4966         (vr_values::extract_range_from_binary_expr): Same.
4967         (vr_values::extract_range_from_unary_expr): Same.
4968         (vr_values::extract_range_from_cond_expr): Same.
4969         (vr_values::extract_range_from_comparison): Same.
4970         (vr_values::check_for_binary_op_overflow): Same.
4971         (vr_values::extract_range_basic): Same.
4972         (vr_values::extract_range_from_assignment): Same.
4973         (compare_ranges): Same.
4974         (compare_range_with_value): Same.
4975         (vr_values::adjust_range_with_scev): Same.
4976         (vrp_valueize): Same.
4977         (vrp_valueize_1): Same.
4978         (vr_values::get_vr_for_comparison): Same.
4979         (vr_values::compare_name_with_value): Same.
4980         (vr_values::compare_names): Same.
4981         (vr_values::vrp_evaluate_conditional): Same.
4982         (find_case_label_ranges): Same.
4983         (vr_values::vrp_visit_switch_stmt): Same.
4984         (vr_values::extract_range_from_phi_node): Same.
4985         (vr_values::simplify_div_or_mod_using_ranges): Same.
4986         (vr_values::simplify_bit_ops_using_ranges): Same.
4987         (test_for_singularity): Same.
4988         (range_fits_type_p): Same.
4989         (vr_values::simplify_cond_using_ranges_1): Same.
4990         (vr_values::simplify_switch_using_ranges): Same.
4991         (vr_values::simplify_float_conversion_using_ranges): Same.
4992         (vr_values::two_valued_val_range_p): Same.
4993         (vr_values::add_equivalence): Move to value_range::equiv_add.
4994         * vr-values.h (vr_values::add_equivalence): Remove.
4995         (VR_INITIALIZER): Remove.
4996         * tree-vrp.c (value_range::set): New.
4997         (value_range::equiv_add): New.
4998         (value_range::value_range): New.
4999         (value_range::deep_copy): New.
5000         (value_range::check): New.
5001         (value_range::equal_p): New.
5002         (value_range::ignore_equivs_equal_p): New.
5003         (value_range::operator==): New.
5004         (value_range::operator!=): New.
5005         (value_range::symbolic_p): New.
5006         (value_range::numeric_p): New.
5007         (value_range::set_undefined): New.
5008         (value_range::set_varying): New.
5009         (value_range::may_contain_p): New.
5010         (value_range::equiv_clear): New.
5011         (value_range::singleton_p): New.
5012         (value_range::intersect): New.
5013         (value_range::dump): New.
5014         (value_range::set_and_canonicalize): New.
5015         (set_value_range): Adjust for value_range API.
5016         (set_value_range_to_undefined): Same.
5017         (set_value_range_to_varying): Same.
5018         (set_and_canonicalize_value_range): Same.
5019         (set_value_range_to_nonnull): Same.
5020         (set_value_range_to_null): Same.
5021         (range_is_null): Same.
5022         (range_is_nonnull): Same.
5023         (range_int_cst_p): Same.
5024         (range_int_cst_singleton_p): Same.
5025         (symbolic_range_p): Same.
5026         (range_includes_zero_p): Same.
5027         (value_range_constant_singleton): Same.
5028         (vrp_set_zero_nonzero_bits): Same.
5029         (ranges_from_anti_range): Same.
5030         (extract_range_into_wide_ints): Same.
5031         (extract_range_from_multiplicative_op): Same.
5032         (set_value_range_with_overflow): Same.
5033         (extract_range_from_binary_expr_1): Same.
5034         (extract_range_from_unary_expr): Same.
5035         (dump_value_range): Same.
5036         (debug_value_range): Same.
5037         (vrp_prop::check_array_ref): Same.
5038         (vrp_prop::check_mem_ref): Same.
5039         (vrp_prop::vrp_initialize): Same.
5040         (vrp_prop::visit_stmt): Same.
5041         (intersect_ranges): Same.
5042         (vrp_prop::visit_phi): Same.
5043         (vrp_prop::vrp_finalize): Same.
5044         (determine_value_range_1): Same.
5045         (determine_value_range): Same.
5046         (vrp_intersect_ranges_1): Rename to...
5047         (vrp_intersect_1): this.
5048         (vrp_intersect_ranges): Rename to...
5049         (value_range::intersect_helper): ...this.
5050         (vrp_meet_1): Rename to...
5051         (value_range::union_helper): ...this.
5052         (vrp_meet): Rename to...
5053         (value_range::union_): ...this.
5054         (copy_value_range): Remove.
5055         * tree-vrp.h (struct value_range): Rewrite into a proper class.
5056         (value_range::vrtype): New.
5057         (value_range::type): New.
5058         (value_range::equiv): New.
5059         (value_range::min): New.
5060         (value_range::max): New.
5061         (value_range::varying_p): New.
5062         (value_range::undefined_p): New.
5063         (value_range::null_p): New.
5064         (value_range::equiv_add): New.
5065         (copy_value_range): Remove.
5067 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
5069         * Makefile.in (SELFTEST_TARGETS): New.
5070         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
5071         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
5072         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
5073         c/Make-lang.in.
5074         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
5075         (selftest-c++-gdb, selftest-c++-valgrind): Move to
5076         cp/Make-lang.in.
5077         * configure: Regenerate.
5078         * configure.ac (selftest_languages): New.
5080 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
5082         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
5083         overflow wraps argument.
5084         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
5085         wide_int_range_multiplicative_op.
5086         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
5087         overflow wraps argument.
5088         (wide_int_range_multiplicative_op): Same.
5089         (wide_int_range_lshift): Same.
5090         (wide_int_range_div): Same.
5091         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
5092         (wide_int_range_lshift): Same.
5093         (wide_int_range_div): Same.
5095 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
5097         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
5098         use sign as argument.
5099         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
5100         wide_int_range_shift_undefined_p.
5102 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5104         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
5105         Rename to...
5106         (@despeculate_copy<ALLI_TI:mode>): ... This.
5107         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
5108         switch statement.
5110 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5112         * config.gcc: Obsolete *-*-solaris2.10*.
5113         * doc/install.texi (Specific, *-*-solaris2*): Document it.
5115 2018-10-12  Jeff Law  <law@redhat.com>
5117         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
5118         reg + sym +- const_int addressing modes.
5120 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
5122         * common.opt (fdiagnostics-minimum-margin-width=): New option.
5123         * diagnostic-show-locus.c (layout::layout): Apply the minimum
5124         margin width.
5125         (layout::start_annotation_line): Only print up to 3 of the
5126         margin character, to avoid touching the left-hand side.
5127         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
5128         minimum margin width, as set by test_diagnostic_context's ctor.
5129         (selftest::test_fixit_insert_containing_newline): Likewise.
5130         (selftest::test_fixit_insert_containing_newline_2): Likewise.
5131         (selftest::test_line_numbers_multiline_range): Clear
5132         dc.min_margin_width.
5133         * diagnostic.c (diagnostic_initialize): Initialize
5134         min_margin_width.
5135         * diagnostic.h (struct diagnostic_context): Add field
5136         "min_margin_width".
5137         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
5138         * opts.c (common_handle_option): Handle
5139         OPT_fdiagnostics_minimum_margin_width_.
5140         * selftest-diagnostic.c
5141         (selftest::test_diagnostic_context::test_diagnostic_context):
5142         Initialize min_margin_width to 6.
5143         * toplev.c (general_init): Initialize global_dc->min_margin_width.
5145 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
5147         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
5148         Fix usage of "error_at_rich_loc" in the comment.
5150 2018-10-15  Renlin Li  <renlin.li@arm.com>
5152         PR target/87563
5153         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
5154         if-conversioned loop when it contains ifn with types not
5155         supported by backend.
5156         * internal-fn.c (expand_direct_optab_fn): Add an assert.
5157         (direct_internal_fn_supported_p): New helper function.
5158         * internal-fn.h (direct_internal_fn_supported_p): Declare.
5160 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
5162         PR target/87572
5163         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5164         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5165         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5167 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
5169         PR tree-optimization/87022
5170         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
5171         bits in dist vector rather than the first one.
5173 2018-10-15  Richard Biener  <rguenther@suse.de>
5175         PR middle-end/87610
5176         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
5177         (visit_loadstore): When a used restrict tag escaped verify that
5178         the points-to solution of "other" pointers do not include
5179         escaped.
5180         (compute_dependence_clique): If a used restrict tag escaped
5181         communicated that down to visit_loadstore.
5183 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
5185         * config/s390/s390.c (s390_expand_vec_init): Force vector element
5186         into reg if it isn't a general operand.
5188 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
5190         PR target/87599
5191         * config/i386/sse.md (*vec_dupv2di): Add register source to
5192         movddup.
5194 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
5196         PR target/87572
5197         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5198         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
5199         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5200         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5202 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
5204         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
5205         (notice_args_size): Set it in the current trace if no insn that can
5206         throw internally has been seen yet.
5207         (connect_traces): When connecting args_size between traces, allow the
5208         incoming values not to match if there is an insn setting it before the
5209         first insn that can throw internally; in that case, force the creation
5210         of a CFI note on this latter insn.
5212 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
5214         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
5215         the base class, because using the injected-class-name was not clearly
5216         specified until DR 176.
5218 2018-10-12  Paul Koning  <ni1d@arrl.net>
5220         * config/pdp11/pdp11.md (doloop_end): New expander.
5221         (doloop_end_insn): renamed from "doloop_end".
5222         (addqi3): New pattern.
5223         (subqi3): New pattern.
5224         * config/pdp11/predicates.md (incdec_operand): New predicate.
5226 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
5228         PR middle-end/81376
5229         * real.c (format_helper::can_represent_integral_type_p): New function
5230         * real.h (format_helper::can_represent_integral_type_p): Ditto.
5231         * match.pd: New pattern.
5233 2018-10-12  Alexandre Oliva <oliva@adacore.com>
5235         * configure.ac: Introduce --enable-large-address-aware
5236         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
5237         * doc/install.texi: Document it.
5238         * configure, config.in: Rebuilt.
5239         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
5240         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
5241         (LINK_SPEC): Insert it.
5242         * config/i386/mingw-w64.h: Likewise.
5244         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
5246 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
5248         PR rtl-optimization/87600
5249         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
5251 2018-10-12  Paul Koning  <ni1d@arrl.net>
5253         * doc/md.texi (doloop_end): Document that the pattern code may
5254         need to check operand mode.
5256 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
5258         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
5259         to zero-extend between int and floating-point registers.
5260         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
5261         ldp into floating-point registers.  Add type and arch attributes.
5262         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
5263         Use f_loads for type attribute.
5265 2018-10-11  Martin Sebor  <msebor@redhat.com>
5267         * doc/extend.texi (attribute packed): Correct typos.
5269 2018-10-11  Martin Sebor  <msebor@redhat.com>
5271         * doc/extend.texi (attribute flatten): Mention interaction with
5272         noinline.
5274 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
5276         PR target/87156
5277         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5278         Set new_decl virtual flag to zero.
5280 2018-10-11  Martin Sebor  <msebor@redhat.com>
5282         PR middle-end/87593
5283         * doc/extend.texi (attribute format_arg): Discuss using multiple
5284         attributes on a single function.
5286 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5288         PR tree-optimization/86829
5289         * match.pd (sin (atan (x))): New simplification rules.
5290         (cos (atan (x))): Likewise.
5291         * real.c (build_sinatan_real): New function.
5292         * real.h (build_sinatan_real): Prototype.
5294 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
5296         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
5297         function.
5298         (fold_mergeeo_helper): New helper function.
5299         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
5300         intrinsics.  Correct some whitespace indentation issues.
5302 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
5304         PR target/87511
5305         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
5306         Use HOST_WIDE_INT_1U for shift.
5308 2018-10-11  Doug Rupp  <rupp@adacore.com>
5309             Olivier Hainque  <hainque@adacore.com>
5311         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
5312         Pass --relax to the linker for RTPs.
5313         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
5315 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
5316             Jan Hubicka  <jh@suse.cz>
5317             Martin Jambor  <mjambor@suse.cz>
5319         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
5320         the same elements are repeated rather than printing all of them.
5321         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
5322         "repeated" elements.
5323         * read-rtl-function.c (test_loading_repeat): New function.
5324         (read_rtl_function_c_tests): Call test_loading_repeat.
5325         * rtl-tests.c (test_dumping_repeat): New function.
5326         (rtl_tests_c_tests): Call test_dumping_repeat.
5328 2018-10-11  Richard Biener  <rguenther@suse.de>
5330         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
5331         bdver?_cost): Unify to ...
5332         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
5333         * config/i386/i386.c (processor_cost_table): Adjust.
5335 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
5337         PR middle-end/87574
5338         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
5339         the thunk when expanding to GIMPLE.
5341 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5343         * varasm.c (mergeable_string_section): Don't try to move zero-length
5344         strings to the merge section.
5346 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
5348         PR target/87573
5349         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
5351 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
5353         PR target/87550
5354         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
5355         to special_args set.
5357 2018-10-10  Richard Biener  <rguenther@suse.de>
5359         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
5360         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
5361         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
5362         and recursing and pattern terminating the recursion on SSE
5363         vector width using ix86_expand_reduc.
5364         (reduc_sminmax_scal_<mode>): Split into part reducing to half
5365         width and recursing and SSE2 vector variant doing the final
5366         reduction with ix86_expand_reduc.
5367         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
5368         with terminating the recursion at AVX level, splitting that
5369         to SSE there.
5371 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
5373         * genmatch.c (error_cb): Rename to...
5374         (diagnostic_cb): ...this, converting int params to enums.
5375         (fatal_at): Update for renaming.
5376         (warning_at): Likewise.
5377         (main): Likewise.
5378         * input.c (selftest::ebcdic_execution_charset::apply):
5379         Update for renaming of...
5380         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
5381         to...
5382         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
5383         converting level and reason to enums.
5384         (class selftest::lexer_error_sink): Rename to...
5385         (class selftest::lexer_test_options): ...this, renaming field
5386         "m_errors" to "m_diagnostics".
5387         (selftest::lexer_test_options::apply): Update for renaming of...
5388         (selftest::lexer_test_options::on_error): ...this, renaming to...
5389         (selftest::lexer_test_options::on_diagnostic): ...this
5390         converting level and reason to enums.
5391         (selftest::test_lexer_string_locations_raw_string_unterminated):
5392         Update for renamings.
5393         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
5394         "reason".
5396 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
5398         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
5399         * config/rs6000/pmmintrin.h: New file.
5401 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
5403         PR tree-optimization/86659
5404         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
5406 2018-10-09  Richard Biener  <rguenther@suse.de>
5408         PR tree-optimization/63155
5409         * tree-ssa-structalias.c: Include tree-ssa.h.
5410         (get_constraint_for_ssa_var): For undefs return nothing_id.
5411         (find_func_aliases): Cleanup PHI handling.
5413 2018-10-09  Richard Biener  <rguenther@suse.de>
5415         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
5416         replacements.
5418 2018-10-09  Martin Liska  <mliska@suse.cz>
5420         * asan.c (asan_emit_stack_protection): If a stack variable
5421         is located in a same file as current function, then emit
5422         line info into variable definition string.
5424 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5426         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
5427         information.
5429 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5431         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
5432         on the thunk.
5434 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5436         PR tree-optimization/86659
5437         * gimple-match.h (struct gimple_match_op): Add reverse field.
5438         (gimple_match_op::set_op): New overloaded method.
5439         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
5440         the REF_REVERSE_STORAGE_ORDER flag on the value.
5441         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
5442         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
5444 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
5446         PR middle-end/63155
5447         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
5448         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
5450 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
5452         PR target/87517
5453         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
5454         Defined with __builtin_ia32_vfmaddsubpd512_mask.
5456 2018-10-08  Richard Biener  <rguenther@suse.de>
5458         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
5459         cost the same as AVX128 ones.
5461 2018-10-08  Paul Koning  <ni1d@arrl.net>
5463         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
5464         (expand_block_move): New function.
5465         * config/pdp11/pdp11.c (output_block_move): Remove.
5466         (expand_block_move): New function.
5467         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
5468         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
5469         (*movmemhi1): Remove.
5471 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
5473         * config/s390/2827.md: Increase latencies for some FP instructions.
5475 2018-10-08  Richard Biener  <rguenther@suse.de>
5477         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5478         Open a dump scope.
5479         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
5480         * tree-vectorizer.h (dump_stmt_cost): Adjust.
5481         (add_stmt_cost): Dump return value of the hook.
5483 2018-10-08  Richard Biener  <rguenther@suse.de>
5485         PR tree-optimization/63155
5486         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
5487         (ssa_propagation_engine::ssa_propagate): Remove redundant
5488         bitmap bit clearing.
5490 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
5492         PR rtl-optimization/86939
5493         PR rtl-optimization/87479
5494         * ira.h (non_conflicting_reg_copy_p): New prototype.
5495         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
5496         (make_hard_regno_dead): Don't add conflicts for register
5497         ignore_reg_for_conflicts.
5498         (make_object_dead): Likewise.
5499         (non_conflicting_reg_copy_p): New function.
5500         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
5501         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
5502         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
5503         (make_hard_regno_dead): Don't add conflicts for register
5504         ignore_reg_for_conflicts.  Remove special conflict handling of
5505         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
5506         check_pic_pseudo_p and update callers.
5507         (mark_pseudo_dead): Don't add conflicts for register
5508         ignore_reg_for_conflicts.
5509         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
5511 2018-10-05  Andrew Waterman  <andrew@sifive.com>
5512             Jim Wilson  <jimw@sifive.com>
5514         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
5515         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
5516         new pattern using HONOR_SNANS that emits one extra instruction.
5518 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
5520         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
5521         patterns): Merge SI and DI patterns to a GPR pattern.
5522         (unnamed define_insn and define_split for record form of that): Merge
5523         to a single define_insn_and_split pattern.
5525 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
5527         PR c++/56856
5528         * input.c (expand_location_to_spelling_point): Add param "aspect"
5529         and use rather than hardcoding LOCATION_ASPECT_CARET.
5530         (get_substring_ranges_for_loc): Handle the case of a single token
5531         within a macro expansion.
5532         * input.h (expand_location_to_spelling_point): Add "aspect" param,
5533         defaulting to LOCATION_ASPECT_CARET.
5535 2018-10-05  Paul Koning  <ni1d@arrl.net>
5537         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
5538         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
5539         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
5540         (pdp11_guard_type): New function.
5542 2018-10-05  Paul Koning  <ni1d@arrl.net>
5544         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
5545         * config/pdp11/pdp11.opt (mfloat32): Remove.
5546         (mfloat64): Remove.
5547         * doc/invoke.texi (pdp11 -mfloat32): Remove:
5548         (pdp11 -mfloat64): Remove.
5550 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
5552         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
5553         (*cmp<mode>_cc_i387): Ditto.
5554         (*cmpu<mode>_cc_i387): Ditto.
5555         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5556         * config/i386/i386.c (ix86_expand_fp_compare): Remove
5557         "scratch" argument.
5558         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
5559         Emit x86_sahf_1 pattern.
5560         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
5561         (ix86_expand_carry_flag_compare): Ditto.
5563 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
5565         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
5566         to reg_or_0_operand.  Add "C" constraint.
5567         (*cmpxf_cc_i387): Ditto.
5568         (*cmp<mode>_i387): Change operand 2 predicate
5569         to nonimm_or_0_operand.  Add "C" constraint.
5570         (*cmp<mode>_cc_i387): Ditto.
5571         (*cmp<mode>_0_i387): Remove insn pattern.
5572         (*cmp<mode>_0_cc_i387): Ditto.
5574 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
5576         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
5577         * config/i386/predicates.md (nonimm_or_0_operand): Rename
5578         from vector_move_operand.  Update all uses.
5580 2018-10-05  Martin Sebor  <msebor@redhat.com>
5582         PR tree-optimization/87490
5583         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
5584         consistently.
5586 2018-10-05  Richard Biener  <rguenther@suse.de>
5588         PR tree-optimization/63155
5589         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
5590         vertical space in dumpfiles.
5591         * tree-ssa-propagate.h
5592         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
5593         * tree-ssa-propagate.c (cfg_blocks_back): New global.
5594         (ssa_edge_worklist_back): Likewise.
5595         (curr_order): Likewise.
5596         (cfg_blocks_get): Remove abstraction.
5597         (cfg_blocks_add): Likewise.
5598         (cfg_blocks_empty_p): Likewise.
5599         (add_ssa_edge): Add to current or next worklist based on
5600         RPO index.
5601         (add_control_edge): Likewise.
5602         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
5603         into ...
5604         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
5605         iteration from CFG and SSA edge worklist so we process
5606         everything in RPO order, prioritizing forward progress
5607         over iteration.
5608         (ssa_prop_init): Allocate new worklists, do not dump
5609         immediate uses.
5610         (ssa_prop_fini): Free new worklists.
5612 2018-10-05  Richard Biener  <rguenther@suse.de>
5614         * tree-core.h (tree_block::abstract_flag): Remove.
5615         (tree_block::block_num): Make full 32bits.
5616         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
5617         * tree.h (BLOCK_ABSTRACT): Remove.
5618         * dwarf2out.c (gen_lexical_block_die): Remove dead code
5619         resulting from BLOCK_ABSTRACT being always false.
5620         (gen_inlined_subroutine_die): Likewise.
5621         (gen_block_die): Likewise.
5622         * tree.c (block_ultimate_origin): Likewise.
5623         * tree-pretty-print.c (dump_block_node): Remove code dealing
5624         with BLOCK_ABSTRACT.
5625         * tree-ssa-live.c (dump_scope_block): Likewise.
5626         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
5627         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
5629 2018-10-05   Richard Biener  <rguenther@suse.de>
5631         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
5632         is asked for initialize mode to the component mode of the
5633         vector type.
5635 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
5637         PR target/87522
5638         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
5639         assembler for -mavx.
5640         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
5642 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
5644         PR target/87509
5645         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
5646         RS6000_BTM_DFP.
5647         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
5648         to be DImode.  When using mffscrn, force the operand to a register.
5650 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
5652         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
5653         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
5654         X87MODEF mode iterator.
5655         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
5656         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
5657         X87MODEF mode iterator.
5659 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
5661         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
5662         -Wno-prio-ctor-dtor.
5664 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5666         * Makefile.in (OBJS): Add opt-problem.o.
5667         * dump-context.h: Include "selftest.h.
5668         (selftest::temp_dump_context): New forward decl.
5669         (class dump_context): Make friend of class
5670         selftest::temp_dump_context.
5671         (dump_context::dump_loc_immediate): New decl.
5672         (class dump_pretty_printer): Move here from dumpfile.c.
5673         (class temp_dump_context): Move to namespace selftest.
5674         (temp_dump_context::temp_dump_context): Add param
5675         "forcibly_enable_dumping".
5676         (selftest::verify_dumped_text):
5677         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
5678         (selftest::verify_item):
5679         (ASSERT_IS_TEXT): Move here from dumpfile.c.
5680         (ASSERT_IS_TREE): Likewise.
5681         (ASSERT_IS_GIMPLE): Likewise.
5682         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
5683         to...
5684         (dump_context::dump_loc_immediate): ...this new function.
5685         (class dump_pretty_printer): Move to dump-context.h.
5686         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5687         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5688         (temp_dump_context::temp_dump_context): Move to "selftest"
5689         namespace.  Add param "forcibly_enable_dumping", and use it to
5690         conditionalize the use of m_pp;
5691         (selftest::verify_dumped_text): Make non-static.
5692         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5693         (selftest::verify_item): Make non-static.
5694         (ASSERT_IS_TEXT): Move to dump-context.h.
5695         (ASSERT_IS_TREE): Likewise.
5696         (ASSERT_IS_GIMPLE): Likewise.
5697         (selftest::test_capture_of_dump_calls): Pass "true" for new
5698         param of temp_dump_context.
5699         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5700         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
5701         TDF_COMPARE_DEBUG.
5702         * opt-problem.cc: New file.
5703         * opt-problem.h: New file.
5704         * optinfo-emit-json.cc
5705         (selftest::test_building_json_from_dump_calls): Pass "true" for
5706         new param of temp_dump_context.
5707         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5708         (optinfo::emit_for_opt_problem): New function.
5709         (optinfo::emit): Clarity which emit_item is used.
5710         * optinfo.h (optinfo::get_dump_location): New accessor.
5711         (optinfo::emit_for_opt_problem): New decl.
5712         (optinfo::emit): Make const.
5713         * selftest-run-tests.c (selftest::run_tests): Call
5714         selftest::opt_problem_cc_tests.
5715         * selftest.h (selftest::opt_problem_cc_tests): New decl.
5716         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5717         bool to opt_result, converting fprintf messages to
5718         opt_result::failure_at calls.  Add "stmt" param for use by the
5719         failure_at calls.
5720         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5721         (runtime_alias_check_p): Convert return type from bool to
5722         opt_result, converting dump_printf calls to
5723         opt_result::failure_at, using the statement DDR_A for their
5724         location.
5725         (find_data_references_in_stmt): Convert return type from bool to
5726         opt_result, converting "return false" to opt_result::failure_at
5727         with a new message.
5728         * tree-data-ref.h: Include "opt-problem.h".
5729         (dr_analyze_innermost): Convert return type from bool to opt_result,
5730         and add a const gimple * param.
5731         (find_data_references_in_stmt): Convert return type from bool to
5732         opt_result.
5733         (runtime_alias_check_p): Likewise.
5734         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5735         dr_analyze_innermost.
5736         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5737         Convert return type from bool to opt_result, adding a message for
5738         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5739         (vect_analyze_data_ref_dependence): Convert return type from bool
5740         to opt_result.  Change sense of return type from "false"
5741         effectively meaning "no problems" to "false" meaning a problem,
5742         so that "return false" becomes "return opt_result::success".
5743         Convert "return true" calls to opt_result::failure_at, using
5744         the location of statement A rather than vect_location.
5745         (vect_analyze_data_ref_dependences): Convert return type from bool
5746         to opt_result.
5747         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5748         calls to opt_result::failure_at, using the stmt location rather
5749         than vect_location.
5750         (vect_verify_datarefs_alignment): Convert return type from bool
5751         to opt_result.
5752         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
5753         into multiple more-tightly-scoped copies.
5754         (vect_analyze_data_refs_alignment): Convert return type from bool
5755         to opt_result.
5756         (vect_analyze_data_ref_accesses): Likewise, converting a
5757         "return false" to a "return opt_result::failure_at", adding a
5758         new message.
5759         (vect_prune_runtime_alias_test_list): Convert return type from
5760         bool to opt_result, converting dump_printf_loc to
5761         opt_result::failure_at.  Add a %G to show the pertinent statement,
5762         and use the stmt's location rather than vect_location.
5763         (vect_find_stmt_data_reference): Convert return type from
5764         bool to opt_result, converting dump_printf_loc to
5765         opt_result::failure_at, using stmt's location.
5766         (vect_analyze_data_refs):  Convert return type from bool to
5767         opt_result.  Convert "return false" to "return
5768         opt_result::failure_at", adding messages as needed.
5769         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5770         type from bool to opt_result.
5771         (vect_determine_vf_for_stmt): Likewise.
5772         (vect_determine_vectorization_factor): Likewise, converting
5773         dump_printf_loc to opt_result::failure_at, using location of phi
5774         rather than vect_location.
5775         (vect_analyze_loop_form_1): Convert return type from bool to
5776         opt_result, converting dump_printf_loc calls, retaining the use of
5777         vect_location.
5778         (vect_analyze_loop_form): Convert return type from loop_vec_info
5779         to opt_loop_vec_info.
5780         (vect_analyze_loop_operations): Convert return type from bool to
5781         opt_result, converting dump_printf_loc calls, using the location
5782         of phi/stmt rather than vect_location where available.  Convert
5783         various "return false" to "return opt_result::failure_at" with
5784         "unsupported phi" messages.
5785         (vect_get_datarefs_in_loop): Convert return type from bool to
5786         opt_result.  Add a message for the
5787         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5788         (vect_analyze_loop_2): Convert return type from bool to
5789         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
5790         each "goto again;", adding new messages where needed.
5791         Add "unsupported grouped {store|load}" messages.
5792         (vect_analyze_loop): Convert return type from loop_vec_info to
5793         opt_loop_vec_info.
5794         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5795         bool to opt_result.
5796         * tree-vect-stmts.c (process_use): Likewise, converting
5797         dump_printf_loc call and using stmt location, rather than
5798         vect_location.
5799         (vect_mark_stmts_to_be_vectorized): Likeise.
5800         (vect_analyze_stmt): Likewise, adding a %G.
5801         (vect_get_vector_types_for_stmt): Convert return type from bool to
5802         opt_result, converting dump_printf_loc calls and using stmt
5803         location, rather than vect_location.
5804         (vect_get_mask_type_for_stmt): Convert return type from tree to
5805         opt_tree, converting dump_printf_loc calls and using stmt location.
5806         * tree-vectorizer.c: Include "opt-problem.h.
5807         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5808         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
5809         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
5810         enabled, use it to report at the top level "couldn't vectorize
5811         loop" followed by the problem.
5812         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5813         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5814         to opt_result.
5815         (vect_analyze_stmt): Likewise.
5816         (vect_get_vector_types_for_stmt): Likewise.
5817         (tree vect_get_mask_type_for_stmt): Likewise.
5818         (vect_analyze_data_ref_dependences): Likewise.
5819         (vect_enhance_data_refs_alignment): Likewise.
5820         (vect_analyze_data_refs_alignment): Likewise.
5821         (vect_verify_datarefs_alignment): Likewise.
5822         (vect_analyze_data_ref_accesses): Likewise.
5823         (vect_prune_runtime_alias_test_list): Likewise.
5824         (vect_find_stmt_data_reference): Likewise.
5825         (vect_analyze_data_refs): Likewise.
5826         (vect_analyze_loop): Convert return type from loop_vec_info to
5827         opt_loop_vec_info.
5828         (vect_analyze_loop_form): Likewise.
5829         (vect_analyze_slp): Convert return type from bool to opt_result.
5831 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5833         * doc/invoke.texi (-fopt-info): Document new "internals"
5834         sub-option.
5835         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5836         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5837         MSG_ALL_KINDS.
5838         (optinfo_verbosity_options): Add "internals".
5839         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5840         (dump_context::apply_dump_filter_p): New member function.
5841         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5842         explicitly masking the dump_kind.
5843         (dump_context::begin_scope): Increment the scope depth first.  Use
5844         apply_dump_filter_p rather than explicitly masking the dump_kind.
5845         (dump_context::emit_item): Use apply_dump_filter_p rather than
5846         explicitly masking the dump_kind.
5847         (dump_dec): Likewise.
5848         (dump_hex): Likewise.
5849         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5850         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5851         (opt_info_switch_p): Update handling of default
5852         MSG_OPTIMIZED_LOCATIONS to cope with default of
5853         MSG_PRIORITY_USER_FACING.
5854         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5855         masking the dump_kind.
5856         (selftest::test_capture_of_dump_calls): Update test_dump_context
5857         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5858         than MSG_ALL.  Generalize scope test to be run at all four
5859         combinations of with/without MSG_PRIORITY_USER_FACING and
5860         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5861         for each of the two values.
5862         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5863         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5864         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5865         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5866         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5867         with MSG_PRIORITY_*.
5868         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5869         dump messages as MSG_PRIORITY_USER_FACING.
5870         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5871         about the interaction with MSG_PRIORITY_*.
5873 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5875         * varasm.c (output_constant): Add new parameter merge_strings.
5876         Make strings properly zero terminated in merge string sections.
5877         (mergeable_string_section): Don't fail if the last char is non-zero.
5878         (assemble_variable_contents): Handle merge string sections.
5879         (assemble_variable): Likewise.
5880         (assemble_constant_contents): Likewise.
5881         (output_constant_def_contents): Likewise.
5882         (output_constructor_array_range,
5883         output_constructor_regular_field): Adjust call to output_constant.
5884         (output_object_block): Adjust call to assemble_constant_contents
5885         and assemble_variable_contents.
5887 2018-10-04  Martin Liska  <mliska@suse.cz>
5889         PR c/87483
5890         * cgraphunit.c (process_function_and_variable_attributes):
5891         Warn about a function with alias attribute and a body.
5893 2018-10-04  Martin Liska  <mliska@suse.cz>
5895         PR ipa/82625
5896         * multiple_target.c (redirect_to_specific_clone): New function.
5897         (ipa_target_clone): Use it.
5898         * tree-inline.c: Fix comment.
5900 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5902         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5903         fields.
5904         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5905         (gcc::dump_manager::register_pass): New member function, adapted
5906         from loop body in gcc::pass_manager::register_pass, adding a
5907         call to update_dfi_for_opt_info.
5908         (gcc::dump_manager::opt_info_enable_passes): Store the
5909         -fopt-info options into the new fields.  Move the loop
5910         bodies into...
5911         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5912         function.
5913         * dumpfile.h (struct opt_pass): New forward decl.
5914         (gcc::dump_manager::register_pass): New decl.
5915         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5916         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5917         "m_optinfo_flags", and "m_optinfo_filename".
5918         * passes.c (gcc::pass_manager::register_pass): Move all of the
5919         dump-handling code to gcc::dump_manager::register_pass.
5921 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5923         PR rtl-optimization/87466
5924         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5925         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5926         * doc/tm.texi: Regenerate.
5927         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5928         * lra-lives.c (process_bb_lives): Likewise.
5929         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5930         Define.
5932 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5934         * params.c (add_params): Fix initialization.
5936 2018-10-04  Martin Liska  <mliska@suse.cz>
5938         PR gcov-profile/84107
5939         * tree-profile.c (init_ic_make_global_vars):
5940         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5941         Come up with new ic_tuple* variables.  Emit
5942         __gcov_indirect_call{,_topn} variables.
5943         (gimple_gen_ic_profiler): Access the variable
5944         and emit gimple.
5945         (gimple_gen_ic_func_profiler): Access
5946         __gcov_indirect_call.callee field.
5947         (gimple_init_gcov_profiler): Use ptr_type_node.
5948         * value-prof.c (gimple_ic): Use ptr_type_node.
5950 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5952         PR tree-optimization/85787
5953         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5954         malloc_candidate_p into this function and add support for
5955         detecting multiple phis.
5956         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5958 2018-10-04  Martin Liska  <mliska@suse.cz>
5960         PR ipa/87491
5961         * ipa-inline.c (inline_to_all_callers_1):
5962         Call ultimate_alias_target for node being inlined.
5964 2018-10-03  Jeff Law  <law@redhat.com>
5966         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5967         target's wchar_t.
5968         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5969         * tree.h (get_typenode_from_name): Prototype.
5971 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5973         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5974         Change operand 2 predicate to nonimmediate_operand.
5975         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5977 2018-10-03  Martin Sebor  <msebor@redhat.com>
5978             Jeff Law  <law@redhat.com>
5980         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5981         initialize it.
5982         (get_string_length): Detect unterminated arrays.
5983         (format_string): Same.
5984         (format_directive): Warn about unterminated arrays.
5985         (handle_gimple_call): Mark statements with no_warning as needed.
5987 2018-10-03  Jim Wilson  <jimw@sifive.com>
5989         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5990         also define __riscv_abi_rve.  Delete trailing white space.
5992 2018-10-03  Paul Koning  <ni1d@arrl.net>
5994         Enable LRA register allocator for PDP11.
5995         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5996         (R): Likewise.
5997         (D): Likewise.
5998         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5999         * config/pdp11/pdp11.opt (-mlra): New option.
6000         * doc/invoke.texi (PDP-11 Options): Document -mlra.
6002 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
6004         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
6005         (*<absneg:code>extend<mode>xf2): Ditto.
6007 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
6009         PR tree-optimization/87415
6010         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
6011         precision fields.
6013 2018-10-02  Jeff Law  <law@redhat.com>
6015         * gimple-fold.c (get_range_strlen): Only set *nonstr when
6016         an unterminated string is discovered.  Bubble up range
6017         even for unterminated strings.
6018         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
6019         indicates the string was not terminated via NONSTR.
6021 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
6023         * tree-vrp.c (extract_range_from_unary_expr): Special case all
6024         pointer conversions.
6025         Do not do anything special for anti-ranges.
6027 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
6029         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
6030         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
6032 2018-10-03  Martin Liska  <mliska@suse.cz>
6034         PR gcov-profile/86109
6035         * coverage.c (coverage_begin_function): Do not
6036         mark lambdas as artificial.
6037         * tree-core.h (struct GTY): Remove tm_clone_flag
6038         and introduce new lambda_function.
6039         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
6041 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6043         PR target/87474
6044         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
6045         P8_VECTOR and VSX are enabled.
6047 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
6049         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
6050         0x3907 as CPU model number.
6052 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
6054         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
6055         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
6056         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
6057         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
6058         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
6059         * config/s390/s390.md: Likewise. Rename also the cpu attribute
6060         value from arch12 to z14.
6062 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
6064         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
6065         (isinfxf2): Ditto.
6066         (isinf<mode>2): Ditto.
6068 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
6070         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
6071         before emitting fxam.  Perform calculations in XFmode.
6073 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
6075         * match.pd (((X /[ex] A) +- B) * A): New transformation.
6077 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
6079         PR middle-end/87319
6080         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
6081         * tree.c (signed_or_unsigned_type_for): Handle complex.
6083 2018-10-02  Jeff Law  <law@redhat.com>
6085         * gimple-fold.c (get_range_strlen): Remove dead code.
6087 2018-10-02  Martin Sebor  <msebor@redhat.com>
6088             Jeff Law  <law@redhat.com>
6090         * builtins.c (unterminated_array): Add new arguments.
6091         If argument is not terminated, bubble up size and exact
6092         state to callers.
6093         (expand_builtin_strnlen): Detect, avoid expanding
6094         and diagnose unterminated arrays.
6095         (c_strlen): Fill in offset of start of unterminated strings.
6096         * builtins.h (unterminated_array): Update prototype.
6098 2018-10-02  Richard Biener  <rguenther@suse.de>
6100         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
6101         of haddv4df, first reduce to SSE width and exploit the fact
6102         that we only need element zero with the reduction result.
6103         (reduc_plus_scal_v2df): Likewise.
6105 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
6107         * dojump.h (do_jump): Delete.
6108         (do_jump_1): Likewise.
6109         (split_comparison): Move around.
6110         * dojump.c (do_jump): Make static.
6111         (do_jump_1): Likewise.
6112         (jumpifnot): Move around.
6113         (jumpifnot_1): Likewise.
6114         (jumpif): Likewise.
6115         (jumpif_1): Likewise.
6116         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
6118 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
6120         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
6121         insns in the delay slot and add_insn_after for the jump insn.
6123 2018-10-02  Richard Biener  <rguenther@suse.de>
6125         * tree-inline.c (expand_call_inline): Use the location of
6126         the callee declaration for the inline-entry marker.
6127         * final.c (notice_source_line): Remove special-casing of
6128         NOTE_INSN_INLINE_ENTRY.
6130 2018-10-01  Carl Love  <cel@us.ibm.com>
6132         PR 69431
6133         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
6134         (__builtin_mtfsb0): New.
6135         (__builtin_mtfsb1): New.
6136         ( __builtin_set_fpscr_rn): New.
6137         (__builtin_set_fpscr_drn): New.
6138         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
6139         (rs6000_expand_set_fpscr_rn_builtin): Add.
6140         (rs6000_expand_set_fpscr_drn_builtin): Add.
6141         (rs6000_expand_builtin): Add case statement entries for
6142         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
6143         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
6144         RS6000_BUILTIN_MFFSL.
6145         (rs6000_init_builtins): Add ftype initialization and def_builtin
6146         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
6147         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
6148         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
6149         rs6000_mffscdrn): Add define_insn.
6150         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
6151         * doc/extend.texi: Add documentation for the builtins.
6153 2018-10-01  Richard Biener  <rguenther@suse.de>
6155         PR tree-optimization/87465
6156         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
6157         causing branch miscounts.
6159 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6161         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
6162         aarch64_option_default_param):  New.
6163         (params.h): Include.
6164         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
6165         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
6166         stack-clash protection validation code.
6168 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6170         * params.c (validate_param): New.
6171         (add_params): Use it.
6172         (set_param_value): Refactor param validation into validate_param.
6173         (diagnostic.h): Include.
6174         * diagnostic.h (diagnostic_ready_p): New.
6176 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6178         * params.c (set_param_value):
6179         Add index of parameter being validated.
6180         * common/common-target.def (option_validate_param): New.
6181         * common/common-targhooks.h (default_option_validate_param): New.
6182         * common/common-targhooks.c (default_option_validate_param): New.
6183         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
6184         * doc/tm.texi: Regenerate.
6186 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6188         PR target/86486
6189         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6190         Add validation for stack-clash parameters and set defaults.
6192 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6194         PR target/86486
6195         * configure.ac: Add stack-clash-protection-guard-size.
6196         * doc/install.texi: Document it.
6197         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
6198         * params.def: Update comment for guard-size.
6199         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
6200         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
6201         * configure: Regenerate.
6203 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6205         PR target/86486
6206         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
6207         STACK_DYNAMIC_OFFSET): New.
6208         * config/aarch64/aarch64.c (aarch64_layout_frame):
6209         Update outgoing args size.
6210         (aarch64_stack_clash_protection_alloca_probe_range,
6211         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
6213 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6215         PR target/86486
6216         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
6217         probe ranges.
6218         * target.def (stack_clash_protection_alloca_probe_range): New.
6219         (stack_clash_protection_final_dynamic_probe): Remove.
6220         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
6221         (default_stack_clash_protection_final_dynamic_probe): Remove.
6222         * targhooks.c: Likewise.
6223         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
6224         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
6225         * doc/tm.texi: Regenerate.
6227 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6229         PR target/86486
6230         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
6231         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
6232         aarch64_clamp_to_uimm12_shift): New.
6233         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
6234         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
6236 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6238         PR target/86486
6239         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
6241 2018-10-01  Jeff Law  <law@redhat.com>
6242             Richard Sandiford <richard.sandiford@linaro.org>
6243             Tamar Christina  <tamar.christina@arm.com>
6245         PR target/86486
6246         * config/aarch64/aarch64.md
6247         (probe_stack_range): Add k (SP) constraint.
6248         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
6249         STACK_CLASH_MAX_UNROLL_PAGES): New.
6250         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
6251         stack probes for stack clash.
6252         (aarch64_allocate_and_probe_stack_space): New.
6253         (aarch64_expand_prologue): Use it.
6254         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
6255         (aarch64_sub_sp): Add emit_move_imm optional param.
6257 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
6259         PR tree-optimization/87261
6260         * match.pd: Remove trailing whitespace.
6261         Add (x & y) | ~(x | y) -> ~(x ^ y),
6262         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
6264 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
6266         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
6267         constraints.
6268         (ashlsi3_insn): Update instruction constraints.
6269         (ashrsi3_insn): Likewise.
6270         (rotrsi3): Likewise.
6271         (add_shift): Likewise.
6272         * config/arc/constraints.md (Csz): New 32 bit constraint. It
6273         avoids placing in the limm field small constants which, otherwise,
6274         could end into a small instruction.
6276 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
6278         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
6279         destination register is not odd-even.
6280         (umaddsidi4_split): Likewise.
6282 2018-10-01  Richard Biener  <rguenther@suse.de>
6284         * tree-inline.c (expand_call_inline): Store origin of fn
6285         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
6286         * tree.c (block_ultimate_origin): Simplify and do some
6287         checking.
6289 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
6291         * config/i386/mmx.md (EMMS): New int iterator.
6292         (emms): New int attribute.
6293         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
6294         EMMS int iterator.  Explicitly declare clobbers.
6295         (mmx_emms): Remove expander.
6296         (mmx_femms): Ditto.
6297         * config/i386/predicates.md (emms_operation): Remove predicate.
6298         (vzeroall_pattern): New predicate.
6299         (vzeroupper_pattern): Rename from vzeroupper_operation.
6300         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
6301         vzeroupper_pattern and vzeroall_pattern predicates.
6303 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
6305         PR rtl-optimization/86939
6306         * ira-lives.c (make_hard_regno_born): Rename from this...
6307         (make_hard_regno_live): ... to this.  Remove update to conflict
6308         information.  Update function comment.
6309         (make_hard_regno_dead): Add conflict information update.  Update
6310         function comment.
6311         (make_object_born): Rename from this...
6312         (make_object_live): ... to this.  Remove update to conflict information.
6313         Update function comment.
6314         (make_object_dead):  Add conflict information update.  Update function
6315         comment.
6316         (mark_pseudo_regno_live): Call make_object_live.
6317         (mark_pseudo_regno_subword_live): Likewise.
6318         (mark_hard_reg_dead): Update function comment.
6319         (mark_hard_reg_live): Call make_hard_regno_live.
6320         (process_bb_node_lives): Likewise.
6321         * lra-lives.c (make_hard_regno_born): Rename from this...
6322         (make_hard_regno_live): ... to this.  Remove update to conflict
6323         information.  Remove now uneeded check_pic_pseudo_p argument.
6324         Update function comment.
6325         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
6326         to conflict information.  Update function comment.
6327         (mark_pseudo_live): Remove update to conflict information.  Update
6328         function comment.
6329         (mark_pseudo_dead): Add conflict information update.
6330         (mark_regno_live): Call make_hard_regno_live.
6331         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
6332         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
6334 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
6336         PR target/87370
6337         * config/i386/i386.c (construct_container): Use TImode for
6338         BLKmode values in 2 integer registers.
6340 2018-09-29  Jeff Law  <law@redhat.com>
6342         * builtins.c (unterminated_array): Pass in c_strlen_data * to
6343         c_strlen rather than just a tree *.
6344         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
6345         Update recursive calls appropriately.  If caller did not provide a
6346         suitable data pointer, create a local one.  When a non-terminated
6347         string is discovered, bubble up information about the string via the
6348         c_strlen_data object.
6349         * builtins.h (c_strlen): Update prototype.
6350         (c_strlen_data): New structure.
6351         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
6352         For a type 2 call, if c_strlen indicates a non-terminated string
6353         use the length of the non-terminated string.
6354         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
6356 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
6358         PR target/87467
6359         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
6360         __m512d type for __A argument rather than __m512.
6362 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
6364         * match.pd (simple_comparison): Don't optimize if either operand is
6365         a function pointer when target needs function pointer canonicalization.
6367 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6369         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
6370         power5 .. power9 to remove indirection.
6371         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
6372         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
6373         ASM_CPU_476_SPEC): Delete.
6374         (ASM_CPU_SPEC): Adjust.
6375         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
6376         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
6378 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6380         * config.in: Delete HAVE_AS_DCI.
6381         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
6382         * config/rs6000/rs6000.h: Ditto.
6383         * configure.ac: Delete HAVE_AS_DCI.
6384         * configure: Regenerate.
6386 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6388         * config.in (HAVE_AS_LWSYNC): Delete.
6389         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
6390         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
6391         do it as a .long .
6392         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
6393         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
6394         as a .long .
6395         * configure.ac: Delete HAVE_AS_LWSYNC.
6396         * configure: Regenerate.
6398 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
6399             Pierre-Marie de Rodat  <derodat@adacore.com>
6401         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
6402         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
6403         (cgraph_node::create_thunk): Add indirect_offset parameter.
6404         (thunk_adjust): Likewise.
6405         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
6406         and initialize the corresponding field with it.
6407         (cgraph_node::dump): Dump indirect_offset field.
6408         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
6409         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
6410         (thunk_adjust): Add indirect_offset parameter and deal with it.
6411         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
6412         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
6413         or if the thunk is external or local.  Fix formatting.  Do not chain
6414         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
6415         if any, in the GIMPLE representation.
6416         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
6417         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
6418         (input_node): Read indirect_offset field.
6419         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
6420         call to thunk_adjust.
6421         * tree-nested.c (struct nesting_info): Add thunk_p field.
6422         (create_nesting_tree): Set it.
6423         (convert_all_function_calls): Copy static chain from targets to thunks.
6424         (finalize_nesting_tree_1): Return early for thunks.
6425         (unnest_nesting_tree_1): Do not finalize thunks.
6426         (gimplify_all_functions): Do not gimplify thunks.
6428 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
6430         * opt-suggestions.c (option_proposer::build_option_suggestions):
6431         Release "option_values".
6433 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
6435         * coverage.c (get_coverage_counts): Convert problem-reporting dump
6436         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
6437         * dumpfile.c (kind_as_string): New function.
6438         (dump_loc): Rather than a hardcoded prefix of "note: ", use
6439         kind_as_string to vary the prefix based on dump_kind.
6440         (selftest::test_capture_of_dump_calls): Update for above.
6442 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
6444         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
6445         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
6447 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
6449         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
6450         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
6451         INVALID_REGNUM instead of FPSR_REG.
6452         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
6453         * config/i386/i386.md: Update comment of FP compares.
6454         (fldenv): Do not clobber FPSR_REG.
6456 2018-09-28  Richard Biener  <rguenther@suse.de>
6458         * tree.h (BLOCK_ORIGIN): New.
6459         * omp-expand.c (grid_expand_target_grid_body): Assign
6460         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
6461         * tree-inline.c (remap_block): Likewise.
6462         * auto-profile.c (get_function_decl_from_block): Simplify
6463         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
6464         * langhooks.c (lhd_print_error_function): Likewise.
6465         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
6466         Likewise.
6467         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6468         * tree.c (block_nonartificial_location): Likewise.
6469         (block_ultimate_origin): Likewise.
6470         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
6471         no longer needed LTO case.
6473 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
6474             Jan Hubicka  <jh@suse.cz>
6475             Martin Jambor  <mjambor@suse.cz>
6477         * simplify-rtx.c (simplify_merge_mask): New function.
6478         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
6479         same masks are used in op1 or op2.
6480         (test_vec_merge): New function.
6481         (test_vector_ops): Call test_vec_merge.
6483 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
6485         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
6486         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
6487         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
6488         (cypress_costs): Set it.
6489         (supersparc_costs): Likewise.
6490         (hypersparc_costs): Likewise.
6491         (leon_cost): Likewise.
6492         (leon3_costs): Likewise.
6493         (sparclet_costs): Likewise.
6494         (ultrasparc_costs): Likewise.
6495         (ultrasparc_costs): Likewise.
6496         (niagara_costs): Likewise.
6497         (niagara2_costs): Likewise.
6498         (niagara3_costs): Likewise.
6499         (niagara4_costs): Likewise.
6500         (niagara7_costs): Likewise.
6501         (m8_costs): Likewise.
6502         (TARGET_CAN_FOLLOW_JUMP): Define.
6503         (pass_work_around_errata::gate): Minor tweak.
6504         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
6505         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
6506         Automaitcally clear MASK_FSMULD mask for V7 processors.
6507         (sparc_can_follow_jump): New static function.
6508         (output_ubranch): Deal with CROSSING_JUMP_P.
6509         (sparc_use_sched_lookahead): Rewrite using switch statement.
6510         (sparc_issue_rate): Reorder.
6511         (sparc_branch_cost): New function.
6513 2018-09-27  Martin Sebor  <msebor@redhat.com>
6515         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
6516         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
6517         (int_fits_type_p): Same.
6519 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
6521         * config/i386/i386.md (FPCR_REG): Remove.
6522         (UNSPEC_FLDCW): Remove.
6523         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
6524         (x86_fldcw_1): Remove insn pattern.
6525         (fnstenv): Do not clobber FPCR_REG.
6526         (fldenv): Ditto.
6527         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
6528         (CALL_USED_REGISTERS): Ditto.
6529         (REG_ALLOC_ORDER): Ditto.
6530         (REG_CLASS_CONTENTS): Ditto.
6531         (HI_REGISTER_NAMES): Ditto.
6532         (ADDITIONAL_REGISTER_NAMES): Use defines instead
6533         of numerical constants.
6534         * config/i386/i386.c (regclass_map): Remove fpsr register.
6535         (dbx_register_map): Ditto.
6536         (dbx64_register_map): Ditto.
6537         (svr4_dbx_register_map): Ditto.
6538         (print_reg): Do not handle FPCR_REG.
6540 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
6542         PR target/87149
6543         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
6544         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
6545         Delete, always treat as true.
6546         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6547         Ditto.  Simplify remaining code.
6548         * config/powerpcspe/powerpcspe.h: Ditto.
6549         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
6550         Simplify remaining code.
6551         (rs6000_expand_builtin): Ditto.
6552         * config/rs6000/rs6000.h: Ditto.
6553         * configure.ac: Ditto.
6554         * configure: Regenerate.
6556 2018-09-27  Martin Liska  <mliska@suse.cz>
6558         * coverage.c (get_coverage_counts): Revert the formatting
6559         of missing profile opt info.
6561 2018-09-27  Richard Biener  <rguenther@suse.de>
6563         PR debug/37801
6564         PR debug/87440
6565         * dwarf2out.c (set_block_origin_self): Do not mark outermost
6566         block as we do not output that.
6567         (gen_inlined_subroutine_die): Elide the originally outermost
6568         block, matching what we do for concrete instances.
6569         (decls_for_scope): Add parameter specifying whether to recurse
6570         to subblocks.
6572 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
6573             Tom de Vries  <tom@codesourcery.com>
6575         PR 82089
6577         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
6578         STORE_FLAG_VALUE == 1.
6580 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
6582         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
6583         constant definitions.
6584         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
6585         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
6586         ("speculation_barrier"): New expander definition.
6588 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
6590         PR gcov-profile/86957
6591         * common.opt: New warning option -Wmissing-profile.
6592         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
6593         * doc/invoke.texi: Document -Wmissing-profile.
6595 2018-09-26  Jim Wilson  <jimw@sifive.com>
6597         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
6598         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
6600 2018-09-26  Martin Sebor  <msebor@redhat.com>
6602         * tree.c (zerop): Change return type to bool.
6603         (integer_zerop, integer_onep, integer_each_onep): Same.
6604         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
6605         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
6606         (real_onep, real_minus_onep, chain_index): Same.
6607         (print_type_hash_statistics, type_list_equal): Same.
6608         * tree.h (zerop): Same.
6609         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
6610         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
6611         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
6612         (real_onep, real_minus_onep, chain_index): Same.
6613         (print_type_hash_statistics, type_list_equal): Same.
6615 2018-09-26  Jim Wilson  <jimw@sifive.com>
6617         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
6619 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
6621         PR target/87414
6622         * config/i386/i386.c: Include debug.h and dwarf2out.h.
6623         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
6624         call.
6626 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
6628         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
6630 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6632         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
6633         and FP_SECOND_SSE_REGS.
6634         (REG_CLASS_NAMES): Ditto.
6635         (REG_CLASS_CONTENTS): Ditto.
6636         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
6637         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
6638         (ix86_preferred_output_reload_class): Ditto.
6639         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
6640         clobber constraint to "=&f".
6641         (fix_truncdi_i387): Ditto.
6642         (lrintxfdi2): Ditto.
6643         (fistdi2_<rounding>): Ditto.
6644         (fpremxf4_i387): Change "=u" constraint to "=f".
6645         (fprem1xf4_i387): Ditto.
6646         (sincosxf3): Ditto.
6647         (fptanxf4_i387): Ditto.
6648         (fxtractxf3_i387): Ditto.
6649         (fscalexf4_i387): Ditto.
6650         (atan2xf3): Change "u" constraint to "f".
6651         (fyl2xxf3_i387): Ditto.
6652         (fyl2xp1xf3_i387): Ditto.
6654 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6656         PR target/87439
6657         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
6658         for removed I387_MASK_PM entity.
6661 2018-09-26  Jeff Law  <law@redhat.com>
6662         Revert
6663         2018-09-26  Alexey Neyman  <stilor@att.net>
6665         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6666         headers are no longer pulled in by <isl/val.h>.
6668 2018-09-26  Richard Biener  <rguenther@suse.de>
6670         PR debug/87443
6671         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
6672         or concrete instance DIE to the tree.  Create abstract origin
6673         attributes also for concrete instances.
6675 2018-09-26  Alexey Neyman  <stilor@att.net>
6677         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6678         headers are no longer pulled in by <isl/val.h>.
6680 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
6682         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
6683         Use new helper functions.
6684         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6685         Use new helper functions.
6686         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6687         aarch_mm_needs_release): New declarations.
6688         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6689         aarch_mm_needs_release): New.
6691 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
6693         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6694         (arm32_output_mi_thunk): Deal with long calls.
6696 2018-09-26  Richard Biener  <rguenther@suse.de>
6698         PR debug/87428
6699         PR debug/87362
6700         * tree-inline.c (expand_call_inline): When the location
6701         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6702         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6703         the inserted BLOCK to make inlined_function_outer_scope_p
6704         recognize it.
6705         * dwarf2out.c (add_call_src_coords_attributes): Do not add
6706         coords for reserved locations.
6708 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6710         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6711         (*call_indirect_nonlocal_sysv<mode>): Ditto.
6712         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6713         (*sibcall_nonlocal_sysv<mode>): Ditto.
6714         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6715         (<bd>_<mode>): Ditto.
6716         (<bd>tf_<mode>): Ditto.
6718 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6720         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6721         control string as a list of templates instead of as C code.
6722         (*altivec_movti): Ditto.
6723         * config/rs6000/darwin.md (movdf_low_di): Ditto.
6725 2018-09-25  Jim Wilson  <jimw@sifive.com>
6727         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6728         when target symbol is weak.
6730 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6732         PR c/87387
6733         * builtins.c (unterminated_array): Simplify.
6734         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
6735         where pointer arithmetic is safe.
6737 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6739         PR target/86987
6740         * config/rs6000/altivec.md (altivec_vspltb): Use
6741         const_0_to_15_operand instead of u5bit_cint_operand.
6742         (*altivec_vspltb_internal): Ditto.
6743         (altivec_vspltb_direct): Ditto.
6744         (altivec_vsplth): Use const_0_to_7_operand instead of
6745         u5bit_cint_operand.
6746         (*altivec_vsplth_internal): Ditto.
6747         (altivec_vsplth_direct): Ditto.
6748         (altivec_vspltw): Use const_0_to_3_operand instead of
6749         u5bit_cint_operand.
6750         (*altivec_vspltw_internal): Ditto.
6751         (altivec_vspltw_direct): Ditto.
6752         (altivec_vspltsf): Ditto.
6753         (*altivec_vspltsf_internal): Ditto.
6754         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6755         various splats with the proper size immediate.  Reorder the various
6756         cases by ascending size of immediate, and put all such together.
6758 2018-09-25  Richard Biener  <rguenther@suse.de>
6760         PR debug/83941
6761         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6762         GC-ification.
6763         (maybe_create_die_with_external_ref): Do not create
6764         DW_TAG_imported_unit here.
6765         (add_abstract_origin_attribute): Handle external BLOCK refs.
6766         (dwarf2out_abstract_function): Simplify LTO case.
6767         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6768         rather than using maybe_create_die_with_external_ref.
6770 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
6772         PR target/71278
6773         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6774         (frndintxf2_mask_pm_i387): Ditto.
6775         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6776         Enable for !flag_trapping_math.
6777         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6778         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6779         Change operand 1 predicate to nonimmediate_operand.
6780         (attr "i387_cw"): Remove mask_pm.
6781         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6782         (enum ix86_entity): Remove I387_MASK_PM.
6783         * config/i386/i386.c (ix86_i387_mode_needed): Do not
6784         handle I387_MASK_PM.
6785         (ix86_mode_needed): Ditto.
6786         (ix86_mode_after): Ditto.
6787         (ix86_mode_entry): Ditto.
6788         (ix86_mode_exit): Ditto.
6789         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6791 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
6793         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6794         to_update_switch_stmts to vNULL instead of calling create on them
6795         immediately.
6797 2018-09-25  Richard Biener  <rguenther@suse.de>
6799         PR tree-optimization/87402
6800         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6801         (visit_phi): Re-instantiate handling of supposed to be VARYING
6802         but non-VARYING backedge value.
6804 2018-09-25  Richard Biener  <rguenther@suse.de>
6806         PR debug/83941
6807         * dwarf2out.c (struct sym_off_pair): New.
6808         (external_die_map): New global.
6809         (lookup_decl_die): When in LTO create DIEs lazily from the
6810         external_die_map.
6811         (lookup_block_die): New function, create DIEs lazily in LTO.
6812         (equate_block_to_die): New function.
6813         (dwarf2out_die_ref_for_decl): During WPA get the association
6814         from the external DIE map.
6815         (dwarf2out_register_external_die): Record mapping into the
6816         external DIE map.
6817         (maybe_create_die_with_external_ref): New function split out from
6818         DIE generation part of old dwarf2out_register_external_die.
6819         (add_abstract_origin_attribute): Do not return the DIE.  When
6820         in LTO reference externals directly.
6821         (dwarf2out_abstract_function): When in LTO ignore calls for
6822         decls with external DIEs (already present abstract instances).
6823         (gen_call_site_die): Adjust.
6824         (add_high_low_attributes): Likewise.
6825         (gen_lexical_block_die): Likewise.
6826         (gen_inlined_subroutine_die): Likewie.
6827         (gen_block_die): Likewise.
6828         (dwarf2out_inline_entry): Likewise.
6829         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6830         DIEs.
6832 2018-09-25  Martin Liska  <mliska@suse.cz>
6834         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6835         integers and not by a float value.
6837 2018-09-25  Martin Liska  <mliska@suse.cz>
6839         PR fortran/87394
6840         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6841         instead of NULL.
6842         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6843         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6844         * godump.c (go_format_type): Remove extra parenthesis.
6846 2018-09-25  Martin Liska  <mliska@suse.cz>
6848         * alias.c (set_dest_equal_p): Remove unused function.
6849         * config/i386/i386.c (def_builtin_pure2): Likewise.
6850         * diagnostic-show-locus.c (class layout): Remove
6851         unused field.
6852         (layout::layout): Likewise here.
6853         * dump-context.h (class temp_dump_context): Likewise.
6854         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6855         (add_AT_loclistsptr): Likewise.
6856         (add_AT_offset): Likewise.
6857         (get_AT_hi_pc): Likewise.
6858         (is_comdat_die): Likewise.
6859         (type_is_enum): Likewise.
6860         (ceiling): Likewise.
6861         (add_AT_vms_delta): Likewise.
6862         (is_class_die): Likewise.
6863         * edit-context.c (class line_event): Remove unused field.
6864         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6865         unused function.
6866         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6867         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6868         (ok_for_base_p_nonstrict): Likewise.
6869         * tree-chrec.c (is_not_constant_evolution): Likewise.
6870         (chrec_fold_poly_cst): Likewise.
6871         * tree-if-conv.c (has_pred_critical_p): Likewise.
6872         * tree-ssa-coalesce.c (print_exprs): Likewise.
6873         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6874         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6875         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6876         (value_range_nonnegative_p): Likewise.
6878 2018-09-25  Martin Liska  <mliska@suse.cz>
6880         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6881         Do not handle "GNU Pascal".
6882         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6883         Likewise.
6884         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6885         from documentation. Likewise.
6886         * dbxout.c (dbxout_range_type): Likewise.
6887         * doc/cpp.texi: Likewise.
6888         * doc/extend.texi: Likewise.
6889         * doc/frontends.texi: Likewise.
6890         * doc/invoke.texi: Remove Pascal entry.
6891         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6892         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6894 2018-09-25  Martin Liska  <mliska@suse.cz>
6896         PR middle-end/86078
6897         * doc/invoke.texi: Document all parameters and remove default
6898         of the parameters.
6900 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6902         PR bootstrap/87417
6903         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6904         contains HOST_WIDE_INTs when computing its size.
6906 2018-09-24  Jim Wilson  <jimw@sifive.com>
6908         PR target/87391
6909         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6910         not TARGET_RVE.
6911         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6913 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6915         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6916         access prev before checking it for NULLness in the
6917         AARCH64_FUSE_CMP_BRANCH case.
6919 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6921         PR target/82699
6922         * config/i386/i386.c (rest_of_insert_endbranch): Set
6923         endbr_queued_at_entrance to true and don't insert ENDBR if
6924         x86_function_profiler will be called.
6925         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6926         is true.
6927         * config/i386/i386.h (machine_function): Add
6928         endbr_queued_at_entrance.
6930 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6932         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6933         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6934         types.
6935         (check_attr_test): Use alternative_mask.
6936         (get_attr_value): Likewise.
6937         (compute_alternative_mask): Use alternative_mask and XWINT.
6938         (make_alternative_compare): Use alternative_mask.
6939         (attr_alt_subset_p): Use XWINT.
6940         (attr_alt_subset_of_compl_p): Likewise.
6941         (attr_alt_intersection): Use alternative_mask and XWINT.
6942         (attr_alt_union): Likewise.
6943         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6944         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6945         (simplify_test_exp): Use alternative_mask and XWINT.
6946         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6947         number calculation to support 64 bits.  Generate code that
6948         checks 64-bit masks.
6949         (main): Use alternative_mask.
6950         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6952 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6954         PR target/80080
6955         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6956         RETURN+USE when returning via %r14.
6958 2018-09-24  Martin Liska  <mliska@suse.cz>
6960         * gcov.c (output_lines): Print colorization legend
6961         for both flag_use_colors and flag_use_hotness_colors.
6962         Reword the help.
6964 2018-09-24  Martin Liska  <mliska@suse.cz>
6966         * coverage.c (get_coverage_counts): Use warning_at
6967         with current_function_decl location. Use %qD in warning
6968         message.
6970 2018-09-24  Martin Liska  <mliska@suse.cz>
6972         * memory-block.h (memory_block_pool::release): Annotate with
6973         valgrind that the memory is not accessible.
6975 2018-09-24  Martin Liska  <mliska@suse.cz>
6977         PR sanitizer/85774
6978         * asan.c: Make asan_handled_variables extern.
6979         * asan.h: Likewise.
6980         * cfgexpand.c (expand_stack_vars): Make sure
6981         a representative is unpoison if another
6982         variable in the partition is handled by
6983         use-after-scope sanitization.
6985 2018-09-24  Richard Biener  <rguenther@suse.de>
6987         PR tree-optimization/63155
6988         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6989         the worklist when the edge of the respective argument isn't
6990         executable.
6992 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6994         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6995         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6996         (MASK_CLASS_P): Update for rename.
6997         (MAYBE_MASK_CLASS_P): Ditto.
6998         (REG_CLASS_NAMES): Update.
6999         (REG_CLASS_CONTENT): Update.
7000         * config/i386/i386.c (regclass_map): Update for MASK_REG
7001         and ALL_MASK_REGS rename.
7002         * config/i386/constraints.md (Yk): Update for rename.
7003         (k): Ditto.
7005 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
7007         * config/i386/i386.h (enum reg_class): Remove
7008         EVEX_SSE_REGS and MOD4_SSE_REGS.
7009         (REG_CLASS_NAMES): Update.
7010         (REG_CLASS_CONTENT): Update.
7011         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
7012         registers as ALL_SSE_REGS.
7013         (ix86_additional_allocno_class_p): Remove.
7014         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
7015         (ix86_register_priority): Lower priority of EVEX SSE registers.
7016         Use IN_RANGE macro where appropriate.
7017         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
7018         AVX-5124VNNIW checks.
7019         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
7020         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
7021         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
7022         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
7023         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
7024         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
7025         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
7026         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
7027         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
7028         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
7029         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
7030         * config/i386/constraints.md (Yh): Remove.
7032 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
7034         * config/i386/i386.c (regclass_map): Declare integer REX registers
7035         as GENERAL_REGS.
7037 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
7039         * doc/service.texi (Service): Switch the fsf.org link to https.
7041 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
7043         PR target/86798
7044         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7045         Define to speculation_safe_value_not_needed.
7047 2018-09-21  Florian Weimer  <fweimer@redhat.com>
7049         PR middle-end/81035
7050         * doc/extend.texi (Common Function Attributes): Mention that
7051         noreturn suppresses tail call optimization.
7053 2018-09-21  Jeff Law  <law@redhat.com>
7055         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
7056         vr_values::cleanup_edges_and_switches.
7057         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
7058         vr_values class.
7059         (identify_jump_threads): Remove EDGE_IGNORE handling.
7060         (execute_vrp): Move handling of to_remove_edges and
7061         to_update_switch_stmts into vr_values class member functions.
7062         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
7063         (to_update_switch_stmts): Likewise.
7064         * vr-values.c: Include cfghooks.h.
7065         (vr_values::vr_values): Initialize to_remove_edges and
7066         to_update_switch_stmts.
7067         (vr_values::~vr_values): Verify to_remove_edges and
7068         to_update_switch_stmts are empty.
7069         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
7070         (vr_values::cleanup_edges_and_switches): New member function.
7071         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
7072         function.  Add new data members.
7074 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
7076         PR tree-optimization/87309
7077         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
7078         calls with pflags and alt_flags.
7079         (selftest::test_capture_of_dump_calls): Add test of interaction of
7080         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
7082 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7084         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
7086 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7088         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
7090 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7092         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
7093         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
7094         (VXWORKS_LIBS_RTP): Minor reordering.
7096 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7098         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
7099         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
7101 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7103         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
7104         (PTRDIFF_TYPE): Likewise.
7106 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7108         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
7109         triplet, similar to support for VxWorks7.
7110         * config/vxworks-dummy.h: Provide a default definition
7111         of TARGET_VXWORKS64 to 0.
7113 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7115         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
7116         * config/vxworks-dummy.h: here.
7118 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7120         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
7122 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
7123             Bo Zhou  <zbo.zhou@hisilicon.com>
7125         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
7126         * config/aarch64/aarch64-tune.md: Regenerated.
7127         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
7128         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
7129         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
7131 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
7132             Julian Brown  <julian@codesourcery.com>
7134         * builtins.c (get_builtin_sync_mem): Handle address spaces.
7136 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
7138         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
7139         if the call takes a static chain.
7141 2018-09-21  Martin Liska  <mliska@suse.cz>
7143         * auto-profile.c (autofdo_source_profile::read): Do not
7144         set sum_all.
7145         (read_profile): Do not add working sets.
7146         (read_autofdo_file): Remove sum_all.
7147         (afdo_callsite_hot_enough_for_early_inline): Remove const
7148         qualifier.
7149         * coverage.c (struct counts_entry): Remove gcov_summary.
7150         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
7151         do not support GCOV_TAG_PROGRAM_SUMMARY.
7152         (get_coverage_counts): Remove summary and expected
7153         arguments.
7154         * coverage.h (get_coverage_counts): Likewise.
7155         * doc/gcov-dump.texi: Remove -w option.
7156         * gcov-dump.c (dump_working_sets): Remove.
7157         (main): Do not support '-w' option.
7158         (print_usage): Likewise.
7159         (tag_summary): Likewise.
7160         * gcov-io.c (gcov_write_summary): Do not dump
7161         histogram.
7162         (gcov_read_summary): Likewise.
7163         (gcov_histo_index): Remove.
7164         (gcov_histogram_merge): Likewise.
7165         (compute_working_sets): Likewise.
7166         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
7167         it not obsolete.
7168         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
7169         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
7170         (GCOV_HISTOGRAM_SIZE): Remove.
7171         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
7172         (struct gcov_summary): Simplify rapidly just
7173         to runs and sum_max fields.
7174         (gcov_histo_index): Remove.
7175         (NUM_GCOV_WORKING_SETS): Likewise.
7176         (compute_working_sets): Likewise.
7177         * gcov-tool.c (print_overlap_usage_message): Remove
7178         trailing empty line.
7179         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
7180         (output_lines): Remove program related line.
7181         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
7182         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
7183         histogram.
7184         (input_profile_summary): Do not read it.
7185         (merge_profile_summaries): And do not merge it.
7186         (input_symtab): Do not call removed function.
7187         * modulo-sched.c (sms_schedule): Do not print sum_max.
7188         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
7189         removed when histogram method was invented.
7190         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
7191         mode.
7192         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
7193         GCOV coding style.
7194         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
7195         and dump selected value.
7196         * profile.c (add_working_set): Remove.
7197         (get_working_sets): Likewise.
7198         (find_working_set): Likewise.
7199         (get_exec_counts): Do not work with working sets.
7200         (read_profile_edge_counts): Do not inform as sum_max is removed.
7201         (compute_branch_probabilities): Likewise.
7202         (compute_value_histograms): Remove argument for call of
7203         get_coverage_counts.
7204         * profile.h: Do not make gcov_summary const.
7206 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
7208         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
7210 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
7212         PR tree-optimization/86990
7213         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
7214         Check that the entire merged store group is made of constants only for
7215         overlapping stores.
7217 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
7219         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
7220         (VTABLE_VERIFICATION_SPEC): Likewise.
7221         (SANITIZER_EARLY_SPEC): Likewise.
7222         (SANITIZER_SPEC): Likewise.
7223         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
7224         * doc/invoke.texi (Link Options): Document -r.
7226 2018-09-20  Richard Biener <rguenther@suse.de>
7228         PR middle-end/87054
7229         * gimplify.c (gimplify_expr): Retain alignment of
7230         addressable lvalue in dereference.
7232 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
7234         PR bootstrap/87013
7235         * configure.ac: Check for .loc is_stmt support.
7236         * configure, config.in: Rebuilt.
7237         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
7238         if not supported.
7240 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
7242         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
7243         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
7244         -misel=no.
7246 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
7248         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7249         VECTOR_OTHER.
7250         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
7251         case VECTOR_OTHER.
7253 2018-09-20  Marek Polacek  <polacek@redhat.com>
7255         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
7257 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
7259         PR tree-optimization/87288
7260         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
7261         into account when determining PEELING_FOR_NITERS.
7263 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
7265         PR tree-optimization/86877
7266         * tree-vect-loop.c (vect_analyze_loop_2): Call
7267         vect_verify_datarefs_alignment.
7269 2018-09-19  Marek Polacek  <polacek@redhat.com>
7271         * doc/invoke.texi: Document -Wclass-conversion.
7273 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
7275         * config/pa/pa.c (pa_adjust_priority): Delete.
7276         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
7278         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
7279         (atomic_storehi): Likewise.
7280         (atomic_storesi): Likewise.
7281         (atomic_loaddi): Restore compare and swap exchange loop code.
7283 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
7285         PR rtl-optimization/86902
7286         * combine.c (try_combine): When changing the CC mode used, don't change
7287         an unrelated mode in other_insn to that new CC mode.
7289 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
7291         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
7292         with %T in place of calls to dump_generic_expr.
7293         (prune_runtime_alias_test_list): Likewise.
7294         (create_runtime_alias_checks): Likewise.
7295         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
7296         (vect_analyze_data_ref_dependence): Likewise.
7297         (vect_slp_analyze_data_ref_dependence): Likewise.
7298         (vect_record_base_alignment): Likewise.  Use %G in place of call
7299         to dump_gimple_stmt.
7300         (vect_compute_data_ref_alignment): Likewise.
7301         (verify_data_ref_alignment): Likewise.
7302         (vect_find_same_alignment_drs): Likewise.
7303         (vect_analyze_group_access_1): Likewise.
7304         (vect_analyze_data_ref_accesses): Likewise.
7305         (dependence_distance_ge_vf): Likewise.
7306         (dump_lower_bound): Likewise.
7307         (vect_prune_runtime_alias_test_list): Likewise.
7308         (vect_find_stmt_data_reference): Likewise.
7309         (vect_analyze_data_refs): Likewise.
7310         (vect_create_addr_base_for_vector_ref): Likewise.
7311         (vect_create_data_ref_ptr): Likewise.
7312         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
7313         (vect_can_advance_ivs_p): Likewise.
7314         (vect_update_ivs_after_vectorizer): Likewise.
7315         (vect_gen_prolog_loop_niters): Likewise.
7316         (vect_prepare_for_masked_peels): Likewise.
7317         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7318         (vect_determine_vectorization_factor): Likewise.
7319         (vect_is_simple_iv_evolution): Likewise.
7320         (vect_analyze_scalar_cycles_1): Likewise.
7321         (vect_analyze_loop_operations): Likewise.
7322         (report_vect_op): Likewise.
7323         (vect_is_slp_reduction): Likewise.
7324         (check_reduction_path): Likewise.
7325         (vect_is_simple_reduction): Likewise.
7326         (vect_create_epilog_for_reduction): Likewise.
7327         (vect_finalize_reduction:): Likewise.
7328         (vectorizable_induction): Likewise.
7329         (vect_transform_loop_stmt): Likewise.
7330         (vect_transform_loop): Likewise.
7331         (optimize_mask_stores): Likewise.
7332         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
7333         (vect_split_statement): Likewise.
7334         (vect_recog_over_widening_pattern): Likewise.
7335         (vect_recog_average_pattern): Likewise.
7336         (vect_determine_min_output_precision_1): Likewise.
7337         (vect_determine_precisions_from_range): Likewise.
7338         (vect_determine_precisions_from_users): Likewise.
7339         (vect_mark_pattern_stmts): Likewise.
7340         (vect_pattern_recog_1): Likewise.
7341         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7342         (vect_record_max_nunits): Likewise.
7343         (vect_build_slp_tree_1): Likewise.
7344         (vect_build_slp_tree_2): Likewise.
7345         (vect_print_slp_tree): Likewise.
7346         (vect_analyze_slp_instance): Likewise.
7347         (vect_detect_hybrid_slp_stmts): Likewise.
7348         (vect_detect_hybrid_slp_1): Likewise.
7349         (vect_slp_analyze_operations): Likewise.
7350         (vect_slp_analyze_bb_1): Likewise.
7351         (vect_transform_slp_perm_load): Likewise.
7352         (vect_schedule_slp_instance): Likewise.
7353         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7354         (vect_mark_stmts_to_be_vectorized): Likewise.
7355         (vect_init_vector_1): Likewise.
7356         (vect_get_vec_def_for_operand): Likewise.
7357         (vect_finish_stmt_generation_1): Likewise.
7358         (vect_check_load_store_mask): Likewise.
7359         (vectorizable_call): Likewise.
7360         (vectorizable_conversion): Likewise.
7361         (vectorizable_operation): Likewise.
7362         (vectorizable_load): Likewise.
7363         (vect_analyze_stmt): Likewise.
7364         (vect_is_simple_use): Likewise.
7365         (vect_get_vector_types_for_stmt): Likewise.
7366         (vect_get_mask_type_for_stmt): Likewise.
7367         * tree-vectorizer.c (increase_alignment): Likewise.
7369 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
7371         * doc/rtl.texi: Adjust vec_select description.
7372         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
7373         non-constant selectors.
7375 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
7377         * config/aarch64/aarch64-protos.h
7378         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
7379         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
7380         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
7381         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
7382         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
7383         (AARCH64_FL_PROFILE): Move index so flags are ordered.
7384         (AARCH64_ISA_RCPC8_4): New flag.
7385         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
7386         to aarch64_offset_9bit_signed_unscaled_p.
7387         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
7388         and use stlur.
7389         * config/aarch64/constraints.md (Ust): New constraint.
7390         * config/aarch64/predicates.md.
7391         (aarch64_9bit_offset_memory_operand): New predicate.
7392         (aarch64_rcpc_memory_operand): New predicate.
7394 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
7396         PR rtl-optimization/87361
7397         * rtlanal.c (nonzero_bits1): Revert accidental change.
7399 2018-09-19  Richard Biener  <rguenther@suse.de>
7401         PR tree-optimization/87349
7402         PR tree-optimization/87342
7403         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
7405 2018-09-18  Marek Polacek  <polacek@redhat.com>
7407         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
7408         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
7410 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
7412         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
7414 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
7416         PR rtl-optimization/86882
7417         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
7419 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
7421         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
7422         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
7424 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
7426         PR other/87353
7427         * doc/invoke.texi (Link Options): Fix formatting and grammar.
7429 2018-09-18  Richard Biener  <rguenther@suse.de>
7431         PR middle-end/63155
7432         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
7433         (compute_samebase_partition_bases): Likewise.
7434         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
7435         (gimple_can_coalesce_p): Simplify.
7437 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
7439         Handle a library implementation of ffs calling __builtin_ffs.
7440         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
7441         (mmix_init_libfuncs): New function: make __builtin_ffs expand
7442         to __ffsdi2.
7444 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
7446         * diagnostic-show-locus.c (class layout_range): Add field
7447         "m_original_idx".
7448         (layout_range::layout_range): Add "original_idx" param and use it
7449         to initialize new field.
7450         (make_range): Use 0 for original_idx.
7451         (layout::layout): Pass in index to calls to
7452         maybe_add_location_range.
7453         (layout::maybe_add_location_range): Add param "original_idx" and
7454         pass it on to layout_range.
7455         (layout::print_any_labels): Pass on range->m_original_idx to
7456         get_text call.
7457         (gcc_rich_location::add_location_if_nearby): Use 0 for
7458         original_idx.
7459         * gcc-rich-location.h (text_range_label::get_text): Update for new
7460         param.
7461         (range_label_for_type_mismatch::get_text): Likewise.
7463 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
7465         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
7467 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
7469         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
7470         format_string_diagnostic_t.
7471         (fmtwarn_n): Likewise.
7472         * substring-locations.c
7473         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
7474         (format_warning_n_va): Convert to...
7475         (format_string_diagnostic_t::emit_warning_n_va): ...this.
7476         (format_warning_va): Convert to...
7477         (format_string_diagnostic_t::emit_warning_va): ...this.
7478         (format_warning_at_substring): Convert to...
7479         (format_string_diagnostic_t::emit_warning): ...this.
7480         (format_warning_at_substring_n): Convert to...
7481         (format_string_diagnostic_t::emit_warning_n): ...this.
7482         * substring-locations.h (class format_string_diagnostic_t): New
7483         class.
7484         (format_warning_va): Convert to
7485         format_string_diagnostic_t::emit_warning_va.
7486         (format_warning_n_va): Convert to
7487         format_string_diagnostic_t::emit_warning_n_va.
7488         (format_warning_at_substring): Convert to
7489         format_string_diagnostic_t::emit_warning.
7490         (format_warning_at_substring_n): Convert to
7491         format_string_diagnostic_t::emit_warning_n.
7493 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
7494             Bernd Schmidt <bernds_cb1@t-online.de>
7496         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
7497         SImode args.
7499 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
7501         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
7502         operand 0 predicate to nonimmediate operand.
7503         (rint<mode>2_frndint): Remove insn pattern.
7504         (rint<mode>2): Change operand 1 predicate to general_operand.
7505         Extend operand 1 to XFmode and generate rintxf2 insn.
7506         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
7507         Do not use X87MODEF mode macro.
7508         (frndintxf2_<rounding>_i387): Rename from
7509         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
7510         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
7511         to XFmode and generate significandxf3 insn.
7513 2018-09-17  Richard Biener  <rguenther@suse.de>
7515         PR tree-optimization/87328
7516         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
7517         visiting unexecutable backedges when not iterating.
7518         (do_rpo_vn): Mark all edges not executable even when not
7519         iterating.
7521 2018-09-17  Martin Jambor  <mjambor@suse.cz>
7523         PR c/63886
7524         * doc/invoke.texi (Warning Options): Likewise.
7526 2018-09-17  Richard Biener  <rguenther@suse.de>
7528         PR tree-optimization/87301
7529         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
7530         clean EH info from leftover copy assignments.
7532 2018-09-17  Martin Liska  <mliska@suse.cz>
7534         PR gcov-profile/85871
7535         * gcov.c (output_intermediate_file): Fix out of bounds
7536         access.
7538 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
7540         * config/arc/arc.c: Object attributes for core4 not reflected
7541         correctly.
7542         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
7543         core3).
7545 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
7547         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
7549 2018-09-17  Martin Liska  <mliska@suse.cz>
7551         * doc/gcov.texi: Document new option --use-hotness-colors.
7552         * gcov.c (struct source_info): Declare new field.
7553         (source_info::source_info): Set default for maximum_count.
7554         (print_usage): Add new -q option.
7555         (process_args): Process it.
7556         (accumulate_line_info): Save src->maximum_count.
7557         (output_line_beginning): Make color line number if
7558         flag_use_hotness_colors is set.
7559         (output_line_details): Pass default argument value.
7560         (output_lines): Pass src->maximum_count.
7562 2018-09-17  Martin Liska  <mliska@suse.cz>
7564         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
7565         Use processor_names table.
7566         * config/i386/i386.c (ix86_default_align): Use
7567         processor_cost_table for alignment values.
7568         (ix86_option_override_internal): Use processor_names.
7569         (ix86_function_specific_print): Likewise.
7570         * config/i386/i386.h (struct processor_costs):
7571         Add alignment values.
7572         (struct ptt): Remove and replace with const char *.
7573         * config/i386/x86-tune-costs.h (struct processor_costs):
7574         Declare default alignments for all costs.
7576 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
7578         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
7579         symbolics or VR_VARYING ranges for ABS_EXPR.
7580         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
7581         when range will wrap.
7583 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
7585         PR middle-end/86864
7586         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
7587         before and after a JUMP_TABLE_DATA.
7589 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
7591         PR middle-end/87188
7592         * dojump.c (do_compare_and_jump): Canonicalize function pointers
7593         when one operand is a function pointer.  Use POINTER_TYPE_P and
7594         FUNC_OR_METHOD_TYPE_P.
7595         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
7596         * fold-const.c (build_range_check): Likewise.
7597         * match.pd (simple_comparison): Likewise.
7599 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
7601         PR c/82967
7602         * spellcheck.c (get_edit_distance_cutoff): New function.
7603         (selftest::test_edit_distance_unit_test_oneway): Rename to...
7604         (selftest::test_get_edit_distance_one_way): ...this.
7605         (selftest::test_get_edit_distance_unit): Rename to...
7606         (selftest::test_get_edit_distance_both_ways): ...this.
7607         (selftest::test_edit_distances): Move tests to this new function,
7608         and test some more pairs of strings.  Update for above renaming.
7609         (selftest::get_old_cutoff): New function.
7610         (selftest::test_get_edit_distance_cutoff): New function.
7611         (selftest::assert_suggested_for): New function.
7612         (ASSERT_SUGGESTED_FOR): New macro.
7613         (selftest::assert_not_suggested_for): New function.
7614         (ASSERT_NOT_SUGGESTED_FOR): New macro.
7615         (selftest::test_suggestions): New function.
7616         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
7617         tests to selftest::test_edit_distances and call it.  Add calls to
7618         selftest::test_get_edit_distance_cutoff and
7619         selftest::test_suggestions.
7620         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
7621         (best_match::consider): Replace hard-coded cutoff calculation with
7622         a call to...
7623         (best_match::get_cutoff): New declaration.
7624         (best_match::get_best_meaningful_candidate): Likewise.
7626 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7628         * builtins.c (fold_builtin_strlen): Remove TODO comment.
7630 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7632         revert:
7633         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7635         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7636         terminated string literal.
7638 2018-09-14  Martin Sebor  <msebor@redhat.com>
7640         * builtins.c (unterminated_array): Handle ARRAY_REF.
7641         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
7642         * builtins.h (unterminated_array): Declare extern.
7643         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
7644         arrays.
7645         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
7646         calls.
7648 2018-09-14  Martin Sebor  <msebor@redhat.com>
7649             Jeff Law  <law@redhat.com>
7651         * builtins.c (unterminated_array): New.
7652         (expand_builtin_strcpy): Adjust.
7653         (expand_builtin_strcpy_args): Detect unterminated arrays.
7654         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
7655         unterminated arrays.
7656         * gimple-fold.h (get_maxval_strlen): Add argument.
7657         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
7659         * gimple-fold.c (get_range_strlen): Add argument.
7660         (get_maxval_strlen): Adjust.
7661         * gimple-fold.h (get_range_strlen): Add argument.
7663 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
7665         * config/i386/movdirintrin.h: Fix copyright year.
7667 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
7669         * reg-stack.c: Include regs.h.
7670         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
7671         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
7672         FIRST_STACK_REG, not DFmode.
7673         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
7674         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
7675         (change stack): Default register mode to the reg_raw_mode of
7676         FIRST_STACK_REG, not DFmode.
7677         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
7678         (*swapxf): Rename from swapxf.
7680 2018-09-14  Carl Love  <cel@us.ibm.com>
7682         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
7683         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7685 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
7687         PR target/87224
7688         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7689         alternatives.
7691 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
7693         PR target/85628
7694         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7696 2018-09-14  Jason Merrill  <jason@redhat.com>
7698         Fix --enable-gather-detailed-mem-stats.
7699         * hash-table.c (hash_table_usage): Change from variable to function.
7700         * hash-table.h: Adjust.
7701         * Makefile.in: Add missing dependencies on hash-table.h.
7703 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7705         PR tree-optimization/87259
7706         PR lto/87283
7707         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7708         execute_cse_reciprocals_1 has tried transforming.
7710 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
7712         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7713         VR_VARYING for PLUS/MINUS_EXPR.
7715 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7717         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7718         formatting.
7720 2018-09-14  Richard Biener  <rguenther@suse.de>
7722         PR middle-end/63155
7723         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7724         bits for the merged partition.
7726 2018-09-13  Martin Sebor  <msebor@redhat.com>
7727             Bernd Edlinger  <bernd.edlinger@hotmail.de>
7729         * builtins.h (c_srlen): Add argument.
7730         * builtins.c (warn_string_no_nul): New function.
7731         (c_strlen): Add argument and use it.  Update recursive calls.
7732         Pass DECL argument to string_constant to get info on non
7733         terminated strings.  Update *NONSTR as needed.
7734         (fold_builtin_strlen): Add argument to calls to c_strlen.
7735         Warn for unterminated arrays.
7736         (warn_string_no_null): Add prototype.
7737         * expr.c (string_constant): Update arguments.  Update recursive
7738         calls appropriately.  Detect missing NUL terminator and outermost
7739         declaration its missing in.
7740         Improve checks for arrays with nonzero lower bound or elements
7741         that are not a single byte.  Simplify offset computation.
7742         Simplify checks for non-NUL terminated strings.
7743         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7744         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7746 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7748         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7749         correctly.
7750         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
7751         argument.  Fix range checks.
7752         * fold-const.h (c_getstr): Adjust protoype.
7753         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7754         string is constant but contains no NUL byte.
7756         * expr.c (string_constant): Adjust function comment.
7757         Remove bogus check for zero termination.
7759         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7761         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7762         (get_constant_size): Don't make STRING_CSTs larger than they are.
7763         (check_string_literal): New check function for STRING_CSTs.
7764         (output_constant): Use it.
7766 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
7768         PR target/86812
7769         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7771 2018-09-13  Richard Biener  <rguenther@suse.de>
7773         PR tree-optimization/87263
7774         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7775         (struct unwind_state): Add max_rpo field.
7776         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7777         Compute max_rpo, the max RPO number a block can be backwards reached
7778         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
7779         separating it from the iterating mode.
7781 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7783         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7784         (rfs_decision): New scheduling decision.
7786 2018-09-13  Richard Biener  <rguenther@suse.de>
7788         PR bootstrap/87134
7789         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7790         (vn_nary_op_insert_pieces_predicated): Do not write useless
7791         valid_dominated_by_p entry outside of the allocated storage.
7793 2018-09-13  Omar Sandoval  <osandov@osandov.com>
7794             Tom de Vries  <tdevries@suse.de>
7796         PR debug/86985
7797         * dwarf2out.c (is_c): New function.
7798         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7800 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
7802         PR target/85628
7803         * config/aarch64/aarch64.md (*aarch64_bfxil):
7804         Define.
7805         * config/aarch64/constraints.md (Ulc): Define.
7806         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7807         Define.
7808         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7809         New function.
7811 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7813         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7814         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7815         aarch64_layout_frame call.
7816         (aarch64_expand_epilogue): Likewise.
7817         (aarch64_initial_elimination_offset): Likewise.
7818         (aarch64_get_separate_components): Likewise.
7819         (aarch64_use_return_insn_p): Likewise.
7820         (aarch64_layout_frame): Remove unneeded check.
7822 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7824         * configure.ac: Only append
7825         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7826         gcc_config_arguments if it was never reconfigured or last reconfigure
7827         was with different arguments.
7828         * configure: Regenerated.
7830 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7831             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7833         PR middle-end/87290
7834         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7835         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7837 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7839         PR tree-optimization/87287
7840         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7841         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7842         * match.pd (X % C == 0): ... here.  New optimization.
7844 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7846         PR middle-end/82853
7847         * expr.h (maybe_optimize_mod_cmp): Declare.
7848         * expr.c (mod_inv): New function.
7849         (maybe_optimize_mod_cmp): New function.
7850         (do_store_flag): Use it.
7851         * cfgexpand.c (expand_gimple_cond): Likewise.
7853 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7854             Julian Brown  <julian@codesourcery.com>
7856         PR middle-end/86336
7857         * gimplify.c (gimplify_scan_omp_clauses): Set
7858         target_firstprivatize_array_bases in OpenACC parallel and kernels
7859         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7860         OpenACC data regions.
7862 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7864         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7865         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7867 2018-09-12  Richard Biener  <rguenther@suse.de>
7869         PR tree-optimization/87280
7870         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7871         edge but unreachable target.
7872         (do_rpo_vn): For conservatively handling a PHI only mark
7873         the backedge executable but not the block reachable.
7875 2018-09-12  Richard Biener  <rguenther@suse.de>
7877         PR tree-optimization/87266
7878         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7879         visited blocks.
7881 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7883         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7884         constants.
7885         ("trunc<BFP:mode><DFP_ALL:mode>2")
7886         ("trunc<DFP_ALL:mode><BFP:mode>2")
7887         ("extend<BFP:mode><DFP_ALL:mode>2")
7888         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7889         according to the target operand type.
7891 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7892             Andreas Krebbel  <krebbel@linux.ibm.com>
7894         PR tree-optimization/86844
7895         * gimple-ssa-store-merging.c
7896         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7897         there are any overlapping stores in between them, make sure they are
7898         also coalesced or we give up completely.
7900 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7902         PR middle-end/87248
7903         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7904         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7906 2018-09-12  Tom de Vries  <tdevries@suse.de>
7908         * common.opt (gdescribe-dies): Add option.
7909         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7910         attribute for artifical and nameless decls.
7911         (dwarf2out_register_external_die): Add description attribute to
7912         external reference die.
7913         (add_desc_attribute): New functions.
7914         (gen_subprogram_die): Add description attribute to
7915         DW_TAG_call_site_parameter.
7916         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7917         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7918         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7919         -gno-describe-dies.
7920         (@item -gdescribe-dies): Add.
7922 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7924         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7925         (extract_range_from_binary_expr_1: Call
7926         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7927         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7928         depend on sign.
7930 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7932         * gimple-ssa-warn-alloca.c
7933         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7934         field for ALLOCA_BOUND_*_LARGE.
7936 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7938         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7940 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7942         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7943         for clobbers.  Remove obsolete comment.
7945 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7947         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7948         mpxchk, mpxld and mpxst types.
7949         (define_attr length_immediate): Remove all processing of mpx types.
7950         (define_attr prefix_0f): Ditto.
7951         (define_attr memory): Ditto.
7953 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7955         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7956         (log<mode>2): Change operand 1 predicate to general_operand.
7957         Extend operand 1 to XFmode and generate logxf3 insn.
7958         (log10<mode>2): Change operand 1 predicate to general_operand.
7959         Extend operand 1 to XFmode and generate log10xf3 insn.
7960         (log2<mode>2): Change operand 1 predicate to general_operand.
7961         Extend operand 1 to XFmode and generate log2xf3 insn.
7962         (fyl2xp1_extend<mode>xf3_i387): Remove.
7963         (log1p<mode>2): Change operand 1 predicate to general_operand.
7964         Extend operand 1 to XFmode and generate log1pxf3 insn.
7965         (fxtract_extend<mode>xf3_i387): Remove.
7966         (logb<mode>2): Change operand 1 predicate to general_operand.
7967         Extend operand 1 to XFmode and generate logbxf3 insn.
7968         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7969         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7970         (significand<mode>2): Change operand 1 predicate to general_operand.
7971         Extend operand 1 to XFmode and generate significandxf3 insn.
7973 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7975         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7976         (load_specs): Use fatal_error.
7977         (DELETE_IF_ORDINARY, process_command): Use error.
7978         (execute, run_attempt): Use fatal_error.
7980 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7982         * diagnostic-core.h (sorry_at): New prototype.
7983         * diagnostic.c (sorry_at): New function.
7985 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7987         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7988         by zero as VR_UNDEFINED.
7990 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7992         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7993         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7994         (<sincos>mode2): New expander.
7995         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7996         (sincos -> sin, cos splitters): Remove splitter patterns.
7997         (sincos<mode>3): Change operand 2 predicate to general_operand.
7998         Extend operand 2 to XFmode and generate sincosxf3 insn.
7999         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
8000         Change operand 3 predicate to const1_operand.
8001         (fptan_extend<mode>xf4_i387): Remove insn pattern.
8002         (tanxf2): Update operands in the call to fptanxf4_i387.
8003         (tan<mode>2): Change operand 1 predicate to general_operand.
8004         Extend operand 1 to XFmode and generate tanxf3 insn.
8005         (atan2xf3): Rename from *fpatanxf3_i387.
8006         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
8007         (atan2xf3): Remove expander.
8008         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
8009         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
8010         (atan<mode>2): Change operand 1 predicate to general_operand.
8011         Extend operand 1 to XFmode and generate atanxf3 insn.
8013 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
8015         * config/i386/i386.md (x87/SSE constant load splitter): Use
8016         memory_operand instead of nonimmediate_operand for input operand
8017         predicate.
8019 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
8021         * config/i386/i386.md (float partial SSE register stall splitter): Move
8022         splitter near its instruction pattern.
8023         (float_extend partial SSE register stall splitter): Ditto.
8024         (float_truncate partial SSE register stall splitter): Ditto.
8026 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
8028         PR target/86794
8029         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8030         to speculation_safe_value_not_needed.
8032         PR target/85666
8033         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
8034         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
8035         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
8036         leaf_function_p, instead use has_hard_reg_initial_val.
8038 2018-09-09  Nathan Sidwell  <nathan@acm.org>
8040         * gcc.h (pfatal_with_name): Don't declare here.
8041         * gcc.c (pfatal_with_name): Make static.
8043 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
8045         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
8046         earlyclobber.
8048 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
8050         PR rtl-optimization/85458
8051         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
8052         priority hook to reduce the priority of EXPR.
8054 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
8056         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
8057         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
8058         with X87_ENABLE_FLOAT.  Remove preparation code.
8059         (*float<SWI48:mode><MODEF:mode>2): Rename from
8060         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
8061         math using "enabled" attribute.
8062         (*floatdi<MODEF:mode>2_i387): Rename from
8063         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
8064         enable for 32bit targets only.
8065         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
8066         splitter.
8067         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
8068         as operand 1 predicate.  Rewrite as define_insn_and_split.
8069         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
8071 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
8073         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
8074         to fallthru to FLOAT case.
8076 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8078         PR target/86731
8079         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
8080         around folding of vec_sl to handle out of range shift values.
8082 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
8084         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
8085         Update callers to gen_fix_trunc<mode>_i387_fisttp
8086         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
8087         nonimmediate_operand.
8088         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
8089         and corresponding splitters.
8090         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
8091         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
8092         (fix_truncdi_i387_with_temp): Remove insn pattern
8093         and corresponding splitters.
8094         (fix_trunc<mode>_i387): Change operand 0 predicate to
8095         nonimmediate_operand.
8096         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
8097         and corresponding splitters.
8098         (*fistdi2_1): Remove.
8099         (fistdi2): Ditto.
8100         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
8101         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
8102         (*fist<mode>2_1): Remove.
8103         (fist<mode>2): Ditto.
8104         (fist<mode>2_with_temp): Remove insn pattern and corresponding
8105         splitters.
8106         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
8107         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
8108         (fistdi2_<rounding>): Change operand 0 predicate to
8109         nonimmediate_operand.
8110         (fistdi2_<rounding>_with_temp): Remove insn pattern
8111         and corresponding splitters.
8112         (fist<mode>2_<rounding>): Change operand 0 predicate to
8113         nonimmediate_operand.
8114         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
8115         and corresponding splitters.
8117         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
8119 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8121         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
8122         the init value.
8124 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8126         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
8127         early gimple folding of vec_splat().
8128         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
8129         * gimple-fold.h: Add an extern define for tree_vec_extract().
8131 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8133         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
8134         wrappers around TREE_TYPE comparisons.
8136 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8138         PR target/80080
8139         * config/s390/predicates.md: Add nonsym_memory_operand.
8140         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
8141         contains a SYMBOL_REF, load it into an intermediate pseudo.
8142         (s390_emit_compare_and_swap): Legitimize operand.
8143         * config/s390/s390.md: Use the new nonsym_memory_operand
8144         with UNSPECV_CAS patterns.
8146 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8148         PR target/80080
8149         * config/s390/s390-passes.def: New file.
8150         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
8151         declaration.
8152         (make_pass_s390_early_mach): Add declaration.
8153         * config/s390/s390.c (make_pass_s390_early_mach):
8154         (s390_option_override): Remove dynamic registration.
8155         * config/s390/t-s390: Add s390-passes.def.
8157 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8159         * config/s390/s390.c (s390_decompose_constant_pool_ref):
8160         Remove UNSPEC_LTREL_BASE check.
8161         (annotate_constant_pool_refs): Likewise.
8162         (find_constant_pool_ref): Likewise.
8163         (find_ltrel_base): Removed.
8164         (replace_ltrel_base): Removed.
8165         (s390_mainpool_finish): Remove replace_ltrel_base call.
8166         (s390_chunkify_start): Remove pending LTREL_BASE logic.
8167         (s390_chunkify_finish): Remove replace_ltrel_base call.
8168         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
8170 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
8172         PR target/86779
8173         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8174         to speculation_safe_value_not_needed.
8176 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
8177             Bernd Schmidt  <bernds_cb1@t-online.de>
8179         * config/nvptx/nvptx-opts.h: New file.
8180         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
8181         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
8182         (ASM_SPEC): Define.
8183         (TARGET_SM35): New macro.
8184         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
8185         correct predicate.
8186         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
8187         values.
8188         (misa=): New option.
8189         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
8191 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
8193         * config/i386/i386.md (truncdfsf2): Remove expander.
8194         (truncdfsf2_with_temp): Ditto.
8195         (truncxf<mode>2): Ditto.
8196         (*truncdfsf_fast_mixed): Remove insn pattern.
8197         (*truncdfsf_fast_i387): Ditto.
8198         (*truncdfsf_mixed): Ditto.
8199         (*truncdfsf_i387): Ditto.
8200         (*truncdfsf2_i387_1): Ditto.
8201         (*truncxfsf2_mixed): Ditto.
8202         (*truncxfdf2_mixed): Ditto.
8203         (*truncxf<mode>2_i387_noop): Ditto. Update callers
8204         to call gen_truncxf<mode>2 instead.
8205         (*truncxf<mode>2_i387): Remove.
8206         (reg->reg splitters): Remove splitter pattern.
8207         (reg->mem splitters): Ditto.
8209         (truncdfsf2): New insn pattern.
8210         (truncxf<mode>2): Ditto.
8212 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8214         * tree-ssa-math-opts.c (is_mult_by): New function.
8215         (is_square_of): Use the above.
8216         (optimize_recip_sqrt): New function.
8217         (pass_cse_reciprocals::execute): Use the above.
8219 2018-09-05  Richard Biener  <rguenther@suse.de>
8221         PR bootstrap/87134
8222         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
8223         to zero-init the emplaced vec.
8225 2018-09-05  Martin Liska  <mliska@suse.cz>
8227         PR tree-optimization/87205
8228         * tree-switch-conversion.c (pass_lower_switch::execute):
8229         Group cases for switch statements.
8231 2018-09-05  Richard Biener  <rguenther@suse.de>
8233         PR tree-optimization/87217
8234         * tree-ssa-sccvn.c (vuse_valueize): New.
8235         (vn_reference_lookup_pieces): Use it.
8236         (vn_reference_lookup): Likewise.
8238 2018-09-05  Nathan Sidwell  <nathan@acm.org>
8240         PR c++/87137
8241         * stor-layout.c (place_field): Scan forwards to check last
8242         bitfield when ms_bitfield_placement is in effect.
8244 2018-09-05  Richard Biener  <rguenther@suse.de>
8246         PR bootstrap/87225
8247         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
8248         return.
8250 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8251             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8253         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
8254         * config.gcc (extra_objs): Build it.
8255         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
8256         Likewise.
8257         * config/aarch64/aarch64-passes.def
8258         (pass_tag_collision_avoidance): New pass.
8259         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
8260         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
8261         (aarch64_classify_address): Remove static qualifier.
8262         (aarch64_address_info, aarch64_address_type): Move to...
8263         * config/aarch64/aarch64-protos.h: ... here.
8264         (make_pass_tag_collision_avoidance): New function.
8265         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
8266         New tuning flag.
8268 2018-09-05  Martin Liska  <mliska@suse.cz>
8270         * doc/gcov.texi: Update documentation of humar
8271         readable mode.
8272         * gcov.c (format_count): Print one decimal place, it provides
8273         more fine number of situations like '1G' vs. '1.4G'.
8275 2018-09-05  Martin Liska  <mliska@suse.cz>
8277         PR target/87164
8278         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
8279         * optc-gen.awk: Allow 'Var' for Deprecated options in order
8280         to generate a MASK value.
8282 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
8284         PR debug/86593
8285         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
8286         if frame pointer isn't used.
8287         (compute_frame_pointer_to_fb_displacement): Likewise.
8289 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
8291         PR target/87198
8292         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
8293         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
8294         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
8295         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
8296         and OPTION_MASK_ISA_XSAVEC_UNSET.
8298 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
8300         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
8301         XOR operations in NAND case.
8303 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
8305         * wide-int-range.cc (wide_int_range_convert): New.
8306         * wide-int-range.h (wide_int_range_convert): New.
8307         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
8308         code into wide_int_range_convert.
8309         (extract_range_into_wide_ints): Do not munge anti range constants
8310         into the entire domain.  Just return the range back.
8312 2018-09-04  Martin Liska  <mliska@suse.cz>
8314         * genmatch.c (output_line_directive): Add new argument
8315         fnargs.
8316         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
8318 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
8320         * doc/invoke.texi (Option Summary): Add whitespace.
8322         * doc/invoke.texi (Option Summary): Add -Waligned-new.
8324 2018-09-04  Richard Biener  <rguenther@suse.de>
8326         PR tree-optimization/87211
8327         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
8328         backedge value we're supposed to treat as VARYING also number
8329         the PHI to VARYING in case it got a different value-number already.
8331 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
8333         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
8334         (extract_range_from_binary_expr_1): Do not call
8335         vrp_can_optimize_bit_op.
8336         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
8337         static.
8338         (wide_int_range_get_mask_and_bounds): New.
8339         (wide_int_range_optimize_bit_op): New.
8340         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
8341         (wide_int_range_bit_and): Same.
8342         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
8343         (wide_int_range_optimize_bit_op): New.
8344         (wide_int_range_get_mask_and_bounds): New.
8346 2018-09-04  Richard Biener  <rguenther@suse.de>
8348         PR tree-optimization/87176
8349         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
8350         variable.  When value-numbering a virtual PHI node make sure
8351         to not value-number to the backedge value.
8353 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
8355         * doc/extend.texi (Long Long, Hex Floats): Document support for
8356         long long and hex floats in more recent versions of ISO C++.
8358 2018-09-03  Richard Biener  <rguenther@suse.de>
8360         PR tree-optimization/87177
8361         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
8362         cleanup.
8364 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8366         * bb-reorder.c (edge_order): Convert to C-qsort-style
8367         tri-state comparator.
8368         (reorder_basic_blocks_simple): Change std::stable_sort to
8369         gcc_stablesort.
8371 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8373         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
8374         tri-state comparator.
8375         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
8377 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8379         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
8380         (mergesort): ... here as maximum count for using netsort.
8381         (gcc_qsort): Set nlim to 3 if stable sort is requested.
8382         (gcc_stablesort): New.
8383         * system.h (gcc_stablesort): Declare.
8385 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8387         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
8388         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
8389         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
8391 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
8393         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
8394         lxsdx and stxsdx alternatives.
8395         (*mov<mode>_hardfloat64): Ditto.
8396         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
8398 2018-09-03  Richard Biener  <rguenther@suse.de>
8400         PR tree-optimization/87200
8401         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
8402         simplify result.
8404 2018-09-03  Martin Liska  <mliska@suse.cz>
8406         PR tree-optimization/87201
8407         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
8408         Fix parenthesis in an expression.
8410 2018-09-03  Richard Biener  <rguenther@suse.de>
8412         PR tree-optimization/87197
8413         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
8414         visited.  CSE the VN_INFO hashtable lookup.
8416         PR tree-optimization/87169
8417         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
8418         iterating make sure there's no extra backedges from irreducible
8419         regions feeding the header.  Mark the destination block
8420         executable.
8422 2018-09-03  Martin Liska  <mliska@suse.cz>
8424         PR driver/83193
8425         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
8426         * common/common-targhooks.c (default_get_valid_option_values):
8427         New function.
8428         * common/common-targhooks.h (default_get_valid_option_values):
8429         Likewise.
8430         * common/config/i386/i386-common.c: Move processor_target_table
8431         from i386.c.
8432         (ix86_get_valid_option_values): New function.
8433         (TARGET_GET_VALID_OPTION_VALUES): New macro.
8434         * config/i386/i386.c (struct ptt): Move to i386-common.c.
8435         (PTA_*): Move all defined masks into i386-common.c.
8436         (ix86_function_specific_restore): Use new processor_cost_table.
8437         * config/i386/i386.h (struct ptt): Moved from i386.c.
8438         (struct pta): Likewise.
8439         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
8440         * doc/tm.texi.in: Likewise.
8441         * opt-suggestions.c (option_proposer::suggest_option):
8442         Pass prefix to build_option_suggestions.
8443         (option_proposer::get_completions): Likewise.
8444         (option_proposer::build_option_suggestions): Use the new target
8445         hook.
8446         * opts.c (struct option_help_tuple): New struct.
8447         (print_filtered_help): Use the new target hook.
8449 2018-09-03  Martin Liska  <mliska@suse.cz>
8451         PR middle-end/59521
8452         * predict.c (set_even_probabilities): Add likely_edges
8453         argument and handle cases where we have precisely one
8454         likely edge.
8455         (combine_predictions_for_bb): Catch also likely_edges.
8456         (tree_predict_by_opcode): Handle gswitch statements.
8457         * tree-cfg.h (find_case_label_for_value): New declaration.
8458         (find_taken_edge_switch_expr): Likewise.
8459         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
8460         Find pivot in decision tree based on probabily, not by number of
8461         nodes.
8463 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
8465         * doc/standards.texi (Standards): Update Objective-C reference.
8467 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8469         * doc/install.texi (Prerequisites): Update link for MPC.
8471 2018-09-01  Michael Matz  <matz@suse.de>
8473         PR tree-optimization/87074
8474         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
8475         PHIs for outer-loop uses.
8477 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8479         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
8480         * doc/invoke.texi (C Dialect Options): Ditto.
8482 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8484         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
8486 2018-08-31  Richard Biener  <rguenther@suse.de>
8488         PR tree-optimization/87168
8489         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
8490         (rpo_elim::eliminate_avail): When OP was not visited it must
8491         be available.
8493 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
8495         * tree-vrp.c (copy_value_range): Convert param "from" from
8496         "value_range *" to "const value_range *".
8497         (range_is_null): Likewise for param "vr".
8498         (range_int_cst_p): Likewise.
8499         (range_int_cst_singleton_p): Likewise.
8500         (symbolic_range_p): Likewise.
8501         (value_ranges_intersect_p): Likewise for both params.
8502         (value_range_nonnegative_p): Likewise for param "vr".
8503         (value_range_constant_singleton): Likewise.
8504         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
8505         (extract_range_into_wide_ints): Likewise for param "vr".
8506         (extract_range_from_multiplicative_op): Likewise for params "vr0"
8507         and "vr1".
8508         (vrp_can_optimize_bit_op): Likewise.
8509         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
8510         "vr1_".
8511         (extract_range_from_unary_expr): Likewise.
8512         (debug_value_range): Likewise for param "vr".
8513         (value_range::dump): Add "const" qualifier.
8514         (vrp_prop::check_array_ref): Convert local "vr" from
8515         "value_range *" to "const value_range *".
8516         (vrp_prop::check_mem_ref): Likewise.
8517         (vrp_prop::visit_stmt): Likewise for local "old_vr".
8518         (vrp_intersect_ranges_1): Likewise for param "vr_1".
8519         (vrp_intersect_ranges): Likewise.
8520         (simplify_stmt_for_jump_threading): Likewise for local "vr".
8521         (vrp_prop::vrp_finalize): Likewise.
8522         * tree-vrp.h (value_range::dump): Add "const" qualifier.
8523         (vrp_intersect_ranges): Add "const" qualifier to params as above.
8524         (extract_range_from_unary_expr): Likewise.
8525         (value_range_constant_singleton): Likewise.
8526         (symbolic_range_p): Likewise.
8527         (copy_value_range): Likewise.
8528         (extract_range_from_binary_expr_1): Likewise.
8529         (range_int_cst_p): Likewise.
8530         (vrp_set_zero_nonzero_bits): Likewise.
8531         (range_int_cst_singleton_p): Likewise.
8533 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
8535         * config/aarch64/arm_neon.h (vabsd_s64): New.
8536         (vnegd_s64): Likewise.
8538 2018-08-31  Martin Jambor  <mjambor@suse.cz>
8540         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
8542 2018-08-31  Martin Liska  <mliska@suse.cz>
8544         * ipa-icf.c (sem_item::add_type): Use
8545         sem_item::m_type_hash_cache.
8546         * ipa-icf.h: Move the cache from sem_item_optimizer
8547         to sem_item.
8549 2018-08-31  Nathan Sidwell  <nathan@acm.org>
8551         * doc/extend.texi (Backwards Compatibility): Remove implicit
8552         extern C leeway of () being (...).
8554 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8556         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
8558 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
8560         PR target/86684
8561         PR target/87149
8562         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
8564 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
8566         PR middle-end/87138
8567         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
8568         gen_int_mode.  Formatting fixes.
8570 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
8572         * target.def (custom_function_descriptors): Improve documentation.
8573         * doc/tm.texi.in (Trampolines): Expand discussion of function
8574         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
8575         beginning of the section.
8576         * doc/tm.texi: Regenerated.
8578 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
8580         * cfg.h (class auto_edge_flag): Spell out the template-id of the
8581         base class in the initializer list.  This is a workaround for
8582         building with older GCC.
8583         (class auto_bb_flag): Likewise.
8585 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
8587         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
8588         (altivec_vcmpequ<VI_char>_p): Remove star.
8589         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
8590         vector load modes.
8591         (expand_strncmp_vec_sequence): New function.
8592         (emit_final_str_compare_vec): New function.
8593         (expand_strn_compare): Add support for vector strncmp.
8594         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
8595         length specification to bytes.
8596         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
8597         (vcmpnezb_p): New pattern.
8598         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
8599         for option -mstring-compare-inline-limit.
8601 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
8603         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
8604         (PTA_SKYLAKE): Add PTA_AES.
8605         (PTA_GOLDMONT): Likewise.
8607 2018-08-29  Jan Hubicka  <jh@suse.cz>
8609         PR lto/86517
8610         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
8611         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
8613 2018-08-29  Jan Hubicka  <jh@suse.cz>
8615         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
8616         TYPE_STUB_DECL.
8617         (hash_tree): Do not visit TYPE_STUB_DECL.
8618         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
8619         stream TYPE_STUB_DECL.
8620         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
8621         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
8622         after free_lang_data.
8623         (type_in_anonymous_namespace_p): Likewise.
8625 2018-08-29  Jan Hubicka  <jh@suse.cz>
8627         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8628         comment that it has to be even number.
8629         (class sreal): Change m_sig type to int32_t.
8630         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8631         int64_t for temporary calculations.
8632         (sreal_verify_basics): Drop one bit from minimum and maximum.
8634 2018-08-30  Richard Biener  <rguenther@suse.de>
8636         PR tree-optimization/87147
8637         * tree-ssa-sccvn.c (SSA_VISITED): New function.
8638         (visit_phi): When the degenerate result is from the backedge and
8639         we didn't visit its definition yet drop to VARYING.
8640         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
8642 2018-08-29  Jan Hubicka  <jh@suse.cz>
8644         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
8645         DECL_VINDEX.
8646         (hash_tree): Likewise.
8648 2018-08-29  Jan Hubicka  <jh@suse.cz>
8650         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
8651         and TYPE_NEXT_REF_TO.
8653 2018-08-29  Jan Hubicka  <jh@suse.cz>
8655         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8656         comment that it has to be even number.
8657         (class sreal): Change m_sig type to int32_t.
8658         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8659         int64_t for temporary calculations.
8660         (sreal_verify_basics): Drop one bit from minimum and maximum.
8662 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
8664         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
8666 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
8668         PR middle-end/86995
8669         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
8670         if to_add is negative.
8672 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8674         PR middle-end/87053
8675         * builtins.c (c_strlen): Improve range checks.
8677 2018-08-29  Martin Sebor  <msebor@redhat.com>
8678             Jeff Law  <law@redhat.com>
8680         PR tree-optimization/86714
8681         PR tree-optimization/86711
8682         * builtins.c (c_strlen): Add arguments to call to string_constant.
8683         * expr.c (string_constant): Add argument.  Detect missing nul
8684         terminator and outermost declaration it's missing in.
8685         * expr.h (string_constant): Add argument.
8686         * fold-const.c (read_from_constant_string): Add arguments to call to
8687         string_constant.
8688         (c_getstr): Likewise.
8689         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8690         to string_constant.
8691         * tree-ssa-strlen.c (get_stridx): Likewise.
8693 2018-08-29  Jan Hubicka  <jh@suse.cz>
8695         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8696         Do not stream DECL_VINDEX.
8697         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8698         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8699         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8701 2018-08-29  Richard Biener  <rguenther@suse.de>
8703         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8704         virtual operands that are not default defs to honor region
8705         boundaries.
8706         (rpo_vn_valueize): Remove ineffective code here.
8708 2018-08-29  Richard Biener  <rguenther@suse.de>
8710         PR tree-optimization/87132
8711         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8712         when skipping defs reachable over backedges.
8714 2018-08-29  Richard Biener  <rguenther@suse.de>
8716         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8717         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8718         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8719         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8720         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8721         (vn_reference_lookup_3): Remove use of const_parms.
8722         (free_rpo_vn): Do not free const_parms.
8723         (do_rpo_vn): Do not call init_const_parms.
8724         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8725         SSA_NAME_POINTS_TO_READONLY_MEMORY.
8726         (call_may_clobber_ref_p_1): Likewise.
8728 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
8730         PR other/86726
8731         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8732         (-O): Ditto.
8733         (-ftree-scev-cprop): Document.
8735 2018-08-29  Jan Hubicka  <jh@suse.cz>
8737         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8738         parameters.
8739         (sreal constructor): Update.
8740         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8741         sreal:operator/): Update.
8743 2018-08-29  Martin Liska  <mliska@suse.cz>
8745         * tree-switch-conversion.c (switch_conversion::expand):
8746         Strenghten assumption about gswitch statements.
8748 2018-08-29  Richard Biener  <rguenther@suse.de>
8750         PR tree-optimization/87117
8751         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8752         re-value-number released SSA VDEFs.
8754 2018-08-29  Richard Biener  <rguenther@suse.de>
8756         PR tree-optimization/87126
8757         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8759 2018-08-28  Jim Wilson  <jimw@sifive.com>
8761         * config/riscv/pic.md: Rewrite.
8762         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8763         invalid address.
8764         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8765         (SOFTF, default_load, softload, softstore): New.
8767 2018-08-28  Jeff Law  <law@redhat.com>
8769         * fold-const.c (fold_binary_loc): Remove recently added assert.
8771 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
8773         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8774         to OP parmeter of generated function.
8776 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
8778         PR tree-optimization/87009
8779         * match.pd: Add boolean optimizations.
8781 2018-08-28  Martin Sebor  <msebor@redhat.com>
8783         PR middle-end/86631
8784         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8785         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8786         (pass_walloca::gate): Use it.
8787         (alloca_call_type): Same.
8788         (pass_walloca::execute): Same.
8789         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8791 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
8793         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8794         GCC_VERSION for usage of "__gcc_dump_printf__" format from
8795         >= 3005 to >= 9000.
8797 2018-08-28  Richard Biener  <rguenther@suse.de>
8799         PR tree-optimization/87124
8800         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8801         constants before looking up avail.
8803 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
8805         PR middle-end/87099
8806         * calls.c (maybe_warn_nonstring_arg): Punt early if
8807         warn_stringop_overflow is zero.  Don't call get_range_strlen
8808         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8809         Swap comparison operands to have constants on rhs.  Only use
8810         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8811         increment lenrng[0].
8813 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8815         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8816         use of tree_to_shwi.  Remove duplicated test for the size being
8817         a whole number of bytes.
8819 2018-08-28  Richard Biener  <rguenther@suse.de>
8821         PR tree-optimization/87117
8822         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8823         Handle removed stmt without LHS (GIMPLE_NOP).
8825 2018-08-28  Richard Biener  <rguenther@suse.de>
8827         PR tree-optimization/87117
8828         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8829         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8831 2018-08-28  Richard Biener  <rguenther@suse.de>
8833         PR tree-optimization/87117
8834         * tree-ssa-pre.c (compute_avail): Do not make expressions
8835         with predicated values available.
8836         (get_expr_value_id): Assert we do not run into predicated value
8837         expressions.
8839 2018-08-28  Richard Biener  <rguenther@suse.de>
8841         PR tree-optimization/87117
8842         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8843         get virtual operands.
8844         (get_expr_operands): Handle STRING_CST like other decls.
8846 2018-08-28  Martin Liska  <mliska@suse.cz>
8848         * tree.h: Update documentation of fndecl_built_in_p
8849         functions.
8852 2018-08-27  Jeff Law  <law@redhat.com>
8853         PR tree-optimization/87110
8854         * tree-ssa-dse.c (compute_trims): Handle non-constant
8855         TYPE_SIZE_UNIT.
8857 2018-08-27  Martin Sebor  <msebor@redhat.com>
8859         PR tree-optimization/86914
8860         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8862 2018-08-27  Martin Sebor  <msebor@redhat.com>
8864         PR tree-optimization/87112
8865         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8866         the type of the bound argument.
8868 2018-08-27  Jeff Law  <law@redhat.com>
8870         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8871         type does not have a TYPE_SIZE_UNIT.
8873 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8875         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8876         with include of backend.h.
8878 2018-08-27  Richard Biener  <rguenther@suse.de>
8880         PR tree-optimization/86927
8881         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8882         use const cond reduction code.
8884 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8886         PR tree-optimization/85758
8887         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8889 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8891         PR c++/87091
8892         * diagnostic-show-locus.c (class layout_range): Update for
8893         conversion of show_caret_p to a tri-state.
8894         (layout_range::layout_range): Likewise.
8895         (make_range): Likewise.
8896         (layout::maybe_add_location_range): Likewise.
8897         (layout::should_print_annotation_line_p): Don't show annotation
8898         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8899         (layout::get_state_at_point): Update for conversion of
8900         show_caret_p to a tri-state.  Bail out early for
8901         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8902         underlining or source colorization.
8903         (gcc_rich_location::add_location_if_nearby): Update for conversion
8904         of show_caret_p to a tri-state.
8905         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8906         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8907         Likewise.
8908         (selftest::test_one_liner_labels): Likewise.
8909         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8910         conversion of show_caret_p to a tri-state.
8911         * pretty-print.c (text_info::set_location): Likewise.
8912         * pretty-print.h (text_info::set_location): Likewise.
8913         * substring-locations.c (format_warning_n_va): Likewise.
8914         * tree-diagnostic.c (default_tree_printer): Likewise.
8915         * tree-pretty-print.c (newline_and_indent): Likewise.
8917 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8919         PR c++/87091
8920         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8921         line above for line-insertion fix-it hints.
8922         (selftest::test_fixit_insert_containing_newline): Update the
8923         expected results, and add a test with line-numbering enabled.
8925 2018-08-27  Martin Liska  <mliska@suse.cz>
8927         PR sanitizer/86962
8928         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8929         params with DECL_HAS_VALUE_EXPR_P.
8931 2018-08-27  Martin Liska  <mliska@suse.cz>
8933         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8934         selected expansion strategy.
8936 2018-08-27  Martin Liska  <mliska@suse.cz>
8938         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8939         * builtins.c (is_builtin_fn): Likewise.
8940         * attribs.c (diag_attr_exclusions): Use new function
8941         fndecl_built_in_p and remove check for FUNCTION_DECL if
8942         possible.
8943         (builtin_mathfn_code): Likewise.
8944         (fold_builtin_expect): Likewise.
8945         (fold_call_expr): Likewise.
8946         (fold_builtin_call_array): Likewise.
8947         (fold_call_stmt): Likewise.
8948         (set_builtin_user_assembler_name): Likewise.
8949         (is_simple_builtin): Likewise.
8950         * calls.c (gimple_alloca_call_p): Likewise.
8951         (maybe_warn_nonstring_arg): Likewise.
8952         * cfgexpand.c (expand_call_stmt): Likewise.
8953         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8954         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8955         (cgraph_node::verify_node): Likewise.
8956         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8957         (cgraph_node::create_clone): Likewise.
8958         * config/arm/arm.c (arm_insert_attributes): Likewise.
8959         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8960         * dse.c (scan_insn): Likewise.
8961         * expr.c (expand_expr_real_1): Likewise.
8962         * fold-const.c (operand_equal_p): Likewise.
8963         (fold_binary_loc): Likewise.
8964         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8965         * gimple-low.c (lower_stmt): Likewise.
8966         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8967         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8968         Likewise.
8969         * gimple.c (gimple_build_call_from_tree): Likewise.
8970         (gimple_call_builtin_p): Likewise.
8971         (gimple_call_combined_fn): Likewise.
8972         * gimplify.c (gimplify_call_expr): Likewise.
8973         (gimple_boolify): Likewise.
8974         (gimplify_modify_expr): Likewise.
8975         (gimplify_addr_expr): Likewise.
8976         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8977         * ipa-cp.c (determine_versionability): Likewise.
8978         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8979         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8980         * ipa-split.c (visit_bb): Likewise.
8981         (split_function): Likewise.
8982         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8983         * lto-cgraph.c (input_node): Likewise.
8984         * lto-streamer-out.c (write_symbol): Likewise.
8985         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8986         (lower_omp_1): Likewise.
8987         * predict.c (strip_predict_hints): Likewise.
8988         * print-tree.c (print_node): Likewise.
8989         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8990         * trans-mem.c (is_tm_irrevocable): Likewise.
8991         (is_tm_load): Likewise.
8992         (is_tm_simple_load): Likewise.
8993         (is_tm_store): Likewise.
8994         (is_tm_simple_store): Likewise.
8995         (is_tm_abort): Likewise.
8996         (tm_region_init_1): Likewise.
8997         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8998         * tree-cfg.c (verify_gimple_call): Likewise.
8999         (move_stmt_r): Likewise.
9000         (stmt_can_terminate_bb_p): Likewise.
9001         * tree-eh.c (lower_eh_constructs_2): Likewise.
9002         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
9003         * tree-inline.c (remap_gimple_stmt): Likewise.
9004         (copy_bb): Likewise.
9005         (estimate_num_insns): Likewise.
9006         (fold_marked_statements): Likewise.
9007         * tree-sra.c (scan_function): Likewise.
9008         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
9009         (optimize_stack_restore): Likewise.
9010         (pass_fold_builtins::execute): Likewise.
9011         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
9012         (mark_all_reaching_defs_necessary_1): Likewise.
9013         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
9014         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
9015         (pass_forwprop::execute): Likewise.
9016         * tree-ssa-loop-im.c (stmt_cost): Likewise.
9017         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
9018         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
9019         * tree-ssa-strlen.c (get_string_length): Likewise.
9020         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
9021         (find_func_aliases_for_call): Likewise.
9022         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
9023         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
9024         * tree-tailcall.c (find_tail_calls): Likewise.
9025         * tree.c (need_assembler_name_p): Likewise.
9026         (free_lang_data_in_decl): Likewise.
9027         (get_call_combined_fn): Likewise.
9028         * ubsan.c (is_ubsan_builtin_p): Likewise.
9029         * varasm.c (incorporeal_function_p): Likewise.
9030         * tree.h (DECL_BUILT_IN): Remove and replace with
9031         fndecl_built_in_p.
9032         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
9033         (fndecl_built_in_p): New.
9035 2018-08-27  Martin Liska  <mliska@suse.cz>
9037         PR tree-optimization/86847
9038         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
9039         Dump also subtree probability.
9040         (switch_decision_tree::do_jump_if_equal): New function.
9041         (switch_decision_tree::emit_case_nodes): Handle special
9042         situations in balanced tree that can be emitted much simpler.
9043         Fix calculation of probabilities that happen in tree expansion.
9044         * tree-switch-conversion.h (struct cluster): Add
9045         is_single_value_p.
9046         (struct simple_cluster): Likewise.
9047         (struct case_tree_node): Add new function has_child.
9048         (do_jump_if_equal): New.
9050 2018-08-27  Martin Liska  <mliska@suse.cz>
9052         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
9053         Add new argument to bit_test_cluster constructor.
9054         (bit_test_cluster::emit): Set bits really number of values
9055         handlel by a test.
9056         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
9057         probability argument.
9058         * tree-switch-conversion.h (struct bit_test_cluster):
9059         Add m_handles_entire_switch.
9061 2018-08-27  Martin Liska  <mliska@suse.cz>
9063         PR tree-optimization/86702
9064         * tree-switch-conversion.c (jump_table_cluster::emit):
9065         Make probabilities even for values in jump table
9066         according to number of cases handled.
9067         (switch_decision_tree::compute_cases_per_edge): Pass
9068         argument to reset_out_edges_aux function.
9069         (switch_decision_tree::analyze_switch_statement): Likewise.
9070         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
9071         Make it static.
9073 2018-08-27  Martin Liska  <mliska@suse.cz>
9075         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
9076         cfun argument explicitly.
9077         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
9078         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
9079         function gimple_switch_default_bb.
9080         (convert_switch_statements):
9081         (expand_builtins):
9082         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
9083         * stmt.c (label_to_block_fn): Use label_to_block and pass
9084         cfun argument explicitly and use gimple_switch_label_bb.
9085         (expand_case): Likewise.
9086         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
9087         cfun argument explicitly. Likewise.
9088         (make_edges_bb): Likewise.
9089         (make_cond_expr_edges): Likewise.
9090         (get_cases_for_edge): Likewise.
9091         (make_gimple_switch_edges): Likewise.
9092         (label_to_block_fn): Likewise.
9093         (label_to_block): Likewise.
9094         (make_goto_expr_edges): Likewise.
9095         (make_gimple_asm_edges): Likewise.
9096         (main_block_label): Likewise.
9097         (group_case_labels_stmt): Likewise.
9098         (find_taken_edge_computed_goto): Likewise.
9099         (find_taken_edge_switch_expr): Likewise.
9100         (gimple_verify_flow_info): Likewise.
9101         (gimple_redirect_edge_and_branch): Likewise.
9102         (gimple_switch_label_bb): New function.
9103         (gimple_switch_default_bb): Likewise.
9104         (gimple_switch_edge): Likewise.
9105         (gimple_switch_default_edge): Likewise.
9106         * tree-cfg.h (label_to_block_fn): Remove and replace ...
9107         (label_to_block): ... with this.
9108         (gimple_switch_label_bb): New.
9109         (gimple_switch_default_bb): Likewise.
9110         (gimple_switch_edge): Likewise.
9111         (gimple_switch_default_edge): Likewise.
9112         * tree-cfgcleanup.c (convert_single_case_switch): Use
9113         new gimple functions and pass new argument to label_to_block.
9114         (cleanup_control_flow_bb):
9115         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
9116         cfun argument explicitly.
9117         (make_eh_edges): Likewise.
9118         (redirect_eh_dispatch_edge): Likewise.
9119         (lower_resx): Likewise.
9120         (lower_eh_dispatch): Likewise.
9121         (maybe_remove_unreachable_handlers): Likewise.
9122         (unsplit_eh): Likewise.
9123         (cleanup_empty_eh): Likewise.
9124         (verify_eh_edges): Likewise.
9125         (verify_eh_dispatch_edge): Likewise.
9126         * tree-ssa-dom.c (record_edge_info): Likewise.
9127         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
9128         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
9129         (thread_through_normal_block): Likewise.
9130         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
9131         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
9132         * tree-switch-conversion.c (switch_conversion::collect): Use new
9133         gimple functions.
9134         (switch_conversion::check_final_bb): Likewise.
9135         (switch_conversion::gather_default_values): Pass new argument
9136         to label_to_block.
9137         (switch_conversion::build_constructors): Likewise.
9138         (switch_decision_tree::compute_cases_per_edge): Use new
9139         gimple_switch_edge function.
9140         (switch_decision_tree::analyze_switch_statement): Pass new argument
9141         to label_to_block.
9142         (switch_decision_tree::try_switch_expansion): Use
9143         gimple_switch_default_edge.
9144         * tree-vrp.c (find_switch_asserts): Pass new argument
9145         to label_to_block.
9146         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
9147         (vr_values::simplify_switch_using_ranges): Likewise.
9149 2018-08-27  Richard Biener  <rguenther@suse.de>
9151         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
9152         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
9154         * tree-ssa-sccvn.h (struct vn_pval): New structure.
9155         (struct vn_nary_op_s): Add unwind_to member.  Add
9156         predicated_values flag and put result into a union together
9157         with a linked list of vn_pval.
9158         (struct vn_ssa_aux): Add name member to make maintaining
9159         a map of SSA name to vn_ssa_aux possible.  Remove no longer
9160         needed info, dfsnum, low, visited, on_sccstack, use_processed
9161         and range_info_anti_range_p members.
9162         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
9163         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
9164         New functions.
9165         (vn_valueize): New global.
9166         (vn_context_bb): Likewise.
9167         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
9168         VN_INFO_PTR_INFO): Remove.
9169         * tree-ssa-sccvn.c: ... (rewrite)
9170         (pass_fre::execute): For -O2+ initialize loops and run
9171         RPO VN in optimistic mode (iterating).  For -O1 and -Og
9172         run RPO VN in non-optimistic mode.
9173         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
9174         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
9175         * doc/invoke.texi (sccvn-max-scc-size): Remove.
9176         (rpo-vn-max-loop-depth): Document.
9177         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
9178         when valuezing the VUSE signals we walked out of the region.
9179         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
9180         (phi_translate): Set VN context block to use for availability
9181         lookup.
9182         (compute_avail): Likewise.
9183         (pre_valueize): New function.
9184         (pass_pre::execute): Adjust to the RPO VN API.
9186         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
9187         (propagate_constants_for_unrolling): Remove.
9188         (tree_unroll_loops_completely): Perform value-numbering
9189         on the unrolled bodies loop parent.
9191 2018-08-27  Richard Biener  <rguenther@suse.de>
9193         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
9194         for partial antic compute.
9196 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
9198         PR rtl-optimization/87065
9199         * combine.c (simplify_if_then_else): Formatting fix.
9200         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
9201         check.
9202         (known_cond): Don't return const_true_rtx for vector modes.  Use
9203         CONST0_RTX instead of const0_rtx.  Formatting fixes.
9205 2018-08-27  Martin Liska  <mliska@suse.cz>
9207         PR gcov-profile/87069
9208         * gcov.c (process_file): Record files already processed
9209         and warn about a file being processed multiple times.
9211 2018-08-27  Martin Liska  <mliska@suse.cz>
9213         PR driver/83193
9214         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9215         Set default values for x_aarch64_*_string strings.
9216         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
9217         prefix.  For -mabi do not print '=ABI' in help and use
9218         <option_value> format for -msve-vector-bits and -moverride
9219         options.
9221 2018-08-26  Jeff Law  <law@redhat.com>
9223         * config/mips/frame-header-opt.c: Include "backend.h" rather than
9224         "cfg.h"
9226 2018-08-26  Marek Polacek  <polacek@redhat.com>
9228         PR c++/87029, Implement -Wredundant-move.
9229         * doc/invoke.texi: Document -Wredundant-move.
9231 2018-08-25  Martin Sebor  <msebor@redhat.com>
9233         PR tree-optimization/87059
9234         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
9235         to the same type as the other.
9236         * fold-const.c (fold_binary_loc): Assert expectation.
9238 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
9240         * config/darwin.c (machopic_legitimize_pic_address): Clean up
9241         extraneous parentheses, dead code section and formatting.
9243 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
9245         PR c++/87091
9246         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
9247         wide enough for jumps in the line-numbering to be visible.
9248         (layout::print_gap_in_line_numbering): New member function.
9249         (layout::calculate_line_spans): When using line numbering, merge
9250         line spans that are only 1 line apart.
9251         (diagnostic_show_locus): When printing line numbers, show gaps in
9252         line numbering directly, rather than printing headers.
9253         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
9254         line-numbering with multiple line spans.
9255         (selftest::test_fixit_insert_containing_newline_2): Add test of
9256         line-numbering, in which the spans are close enough to be merged.
9258 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
9260         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
9261         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
9262         * tree-vrp.c (range_is_nonnull): Remove.
9263         (range_includes_zero_p): Accept value_range instead of min/max.
9264         (extract_range_from_binary_expr_1): Do not early bail on
9265         POINTER_PLUS_EXPR.
9266         Use range_includes_zero_p instead of range_is_nonnull.
9267         (extract_range_from_unary_expr): Use range_includes_zero_p instead
9268         of range_is_nonnull.
9269         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
9270         special case VR_ANTI_RANGE.
9271         (vrp_finalize): Same.
9272         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
9273         instead of min/max.
9274         (range_is_nonnull): Remove.
9275         * vr-values.c (vrp_stmt_computes_nonzero): Use
9276         range_includes_zero_p instead of range_is_nonnull.
9277         (extract_range_basic): Pass value_range to range_includes_zero_p
9278         instead of range_is_nonnull.
9280 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
9282         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
9283         * emit-rtl.h (rtl_data): Remove return_bnd.
9284         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
9285         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
9286         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
9287         (POINTER_BOUNDS_MODE): Remove definition.
9288         (make_pointer_bounds_mode): Remove.
9289         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
9290         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
9291         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
9292         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
9293         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
9294         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
9295         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
9297         * config/i386/i386-modes.def (BND32, BND64): Remove.
9298         * config/i386/i386.c (dbx_register_map): Remove bound registers.
9299         (dbx64_register_map): Ditto.
9300         (svr4_dbx_register_map): Ditto.
9301         (indirect_thunk_bnd_needed): Remove.
9302         (indirect_thunks_bnd_used): Ditto.
9303         (indirect_return_bnd_needed): Ditto.
9304         (indirect_return_via_cx_bnd): Ditto.
9305         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
9306         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
9307         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
9308         (output_indirect_thunk_function): Remove handling of
9309         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
9310         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
9311         (ix86_save_reg): Remove handling of crtl->return_bnd.
9312         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
9313         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
9314         and UNSPEC_BNDLX_ADDR.
9315         (ix86_output_indirect_branch_via_reg): Remove handling of
9316         indirect_thunk_prefix_bnd.
9317         (ix86_output_indirect_branch_via_push): Ditto.
9318         (ix86_output_function_return): Ditto.
9319         (ix86_output_indirect_function_return): Ditto.
9320         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
9321         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
9322         (CALL_USED_REGISTERS): Ditto.
9323         (REG_ALLOC_ORDER): Update for removal of bound registers.
9324         (HI_REGISTER_NAMES): Ditto.
9325         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
9326         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
9327         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
9328         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
9329         (FIRST_PSEUDO_REG): Update.
9330         (BND): Remove mode iterator.
9331         * config/i386/predicates.md (bnd_mem_operator): Remove.
9333 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
9335         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
9336         vectors.
9338 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
9340         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
9341         the case in which the permute needs only a single element and
9342         repeats for every vector of the result.  Extend that case to
9343         handle variable-length vectors.
9344         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9346 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
9348         PR debug/79342
9349         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
9350         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
9352 2018-08-24  Richard Biener  <rguenther@suse.de>
9354         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
9355         bb_flags_allocated members.
9356         (auto_flag): New RAII class for allocating flags.
9357         (auto_edge_flag): New RAII class for allocating edge flags.
9358         (auto_bb_flag): New RAII class for allocating bb flags.
9359         * cfgloop.c (verify_loop_structure): Allocate temporary edge
9360         flag dynamically.
9361         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
9362         in favor of temporarily allocated BB flag.
9363         * hsa-brig.c: Re-order includes.
9364         * hsa-dump.c: Likewise.
9365         * hsa-regalloc.c: Likewise.
9366         * print-rtl.c: Likewise.
9367         * profile-count.c: Likewise.
9369 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
9371         PR target/86989
9372         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
9373         the TOC register.
9375 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
9377         PR 87073/bootstrap
9378         * wide-int-range.cc (wide_int_range_div): Do not ignore result
9379         from wide_int_range_multiplicative_op.
9381 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9383         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
9384         "permutaion".
9386 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
9388         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
9389         to 'expanded'.
9391 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
9393         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
9394         full GENERIC expression used for replacement.
9396 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
9398         * tree-vrp.c (abs_extent_range): Remove.
9399         (extract_range_into_wide_ints): Pass wide ints by reference.
9400         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
9401         Pass wide ints by reference in all calls to
9402         extract_range_into_wide_ints.
9403         * wide-int-range.cc (wide_int_range_div): New.
9404         * wide-int-range.h (wide_int_range_div): New.
9405         (wide_int_range_includes_zero_p): New.
9406         (wide_int_range_zero_p): New.
9408 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
9410         * config/aarch64/aarch64.md (arches): New enum.
9411         (arch): New enum attr.
9412         (arch_enabled): New attr.
9413         (enabled): Now uses arch_enabled only.
9414         (simd, sve, fp16): Removed attribute.
9415         (fp): Attr now defined in terms of 'arch'.
9416         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
9417         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
9418         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
9419         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
9420         attributes into 'arch'.
9421         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
9422         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
9423         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
9424         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
9425         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
9426         'simd' attribute into 'arch'.
9427         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
9428         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
9429         Convert use of 'fp' attribute to 'arch'.
9430         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
9431         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
9432         into 'arch'.
9433         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
9434         (different modes) Merge 'fp' and 'simd' into 'arch'.
9435         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
9436         'simd' into 'arch'.
9438 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
9440         PR rtl-optimization/87026
9441         * expmed.c (canonicalize_comparison): If we can no longer create
9442         pseudoregisters, don't.
9444 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
9446         PR target/86951
9447         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
9448         prototype.
9449         * config/arm/arm.c (speculation_barrier_libfunc): New static
9450         variable.
9451         (arm_init_libfuncs): Initialize it.
9452         (arm_emit_speculation_barrier): New function.
9453         * config/arm/arm.md (speculation_barrier): Call
9454         arm_emit_speculation_barrier for architectures that do not have
9455         DSB or ISB.
9456         (speculation_barrier_insn): Only match on Armv7 or later.
9458 2018-08-23  Richard Biener  <rguenther@suse.de>
9460         PR middle-end/87024
9461         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
9462         calls.
9464 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
9466         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
9467         of single-vector TBLs.
9468         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
9469         one input is given.
9471 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
9473         PR target/85910
9474         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
9475         aarch64_evpc_tbl guard.
9477 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9479         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
9480         behaviour.
9482 2018-08-22  Martin Sebor  <msebor@redhat.com>
9484         PR middle-end/87052
9485         * tree-pretty-print.c (pretty_print_string): Add argument.
9486         (dump_generic_node): Call to pretty_print_string with string size.
9488 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
9490         PR rtl-optimization/86771
9491         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
9492         of two SETs into those two SETs, one to be placed at i2, if that SETs
9493         destination is modified between i2 and i3.
9495 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9497         PR tree-optimization/86725
9498         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
9499         function.
9500         (vect_analyze_scalar_cycles_1): Check it.
9502 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9504         PR tree-optimization/86725
9505         * tree-vect-loop.c (vect_is_simple_reduction): When treating
9506         an outer loop phi as a double reduction, make sure that the
9507         single user of the phi result is an inner loop phi.
9509 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9511         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
9512         grouped stores with gaps to a strided group.
9514 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9516         * tree-vect-stmts.c (get_group_load_store_type)
9517         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
9518         first statement in a group.
9520 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
9522         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
9523         the sequence used in gcc/gcc.c.
9525 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
9527         PR other/704
9528         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
9529         building it.
9531 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
9533         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
9534         Darwin10-specific unwinder-shim.
9535         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
9536         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
9537         New to cater for Darwin10 Rosetta.
9539 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
9541         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
9542         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
9544 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
9546         PR bootstrap/81033
9547         PR target/81733
9548         PR target/52795
9549         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
9550         (dwarf2out_switch_text_section): Generate a local label for the
9551         second function sub-section and apply it as the second FDE start
9552         label.
9553         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
9554         second sub-section start.
9556 2018-08-22  Richard Biener  <rguenther@suse.de>
9558         PR tree-optimization/86988
9559         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
9561 2018-08-22  Richard Biener  <rguenther@suse.de>
9563         PR tree-optimization/86945
9564         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
9566 2018-08-22  Alexandre Oliva <oliva@adacore.com>
9568         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
9569         a comment about how uses of r2 for .sdata2 come about.
9571 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
9573         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
9575 2018-08-21  Marek Polacek  <polacek@redhat.com>
9577         PR c++/86981, Implement -Wpessimizing-move.
9578         * doc/invoke.texi: Document -Wpessimizing-move.
9580 2018-08-21  Jan Hubicka  <jh@suse.cz>
9582         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
9583         * tree.h (is_redundant_typedef): Remove.
9584         * dwarf2out.c (is_redundant_typedef): Turn into static function.
9586 2018-08-21  Jan Hubicka  <jh@suse.cz>
9588         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
9589         when possible.
9591 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
9593         * expmed.c (extract_low_bits): Reject invalid subregs early.
9595 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9597         PR middle-end/86121
9598         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
9599         behaviour.
9601 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9603         * config/vxworks.h: Guard vxworks_asm_out_constructor and
9604         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
9605         * config/vxworks.c: Likewise.
9607 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9609         * config/vxworks.c: Set targetm.have_ctors_dtors
9610         if HAVE_INITFINI_ARRAY_SUPPORT.
9611         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
9612         if HAVE_INITFINI_ARRAY_SUPPORT.
9614 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9616         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
9617         default search path for VxWorks < 7.
9619 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9621         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
9622         (get_format_string): Refer to c_getstr.
9624 2018-08-21  Tom de Vries  <tdevries@suse.de>
9626         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9627         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9628         (debuginfo_early_stop): Declare.
9629         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9630         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9631         (debuginfo_early_stop): New function.
9632         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
9633         and debuginfo_early_stop.
9634         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
9635         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
9636         (general_init): Call debuginfo_early_init.
9637         (finalize): Call debuginfo_fini.
9638         (do_compile): Call debuginfo_init.
9639         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
9640         -fdump-early-debug.
9641         (@item -fdump-debug, @item -fdump-earlydebug): Add.
9643 2018-08-21  Tom de Vries  <tdevries@suse.de>
9645         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
9646         flag_dump_noaddr and flag_dump_unnumbered.
9648 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
9650         * wide-int-range.cc (wide_int_range_abs): New.
9651         (wide_int_range_order_set): Rename from wide_int_range_min_max.
9652         * wide-int-range.h (wide_int_range_abs): New.
9653         (wide_int_range_min_max): New.
9654         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
9655         case to call wide_int_range_abs.
9656         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
9657         (extract_range_from_abs_expr): Delete.
9659 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
9661         PR target/87033
9662         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
9663         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
9664         for indexed loads.
9666 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9667             Jeff Law <law@redhat.com>
9669         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
9670         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
9672 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
9674         PR other/84889
9675         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
9676         (decl_attributes): Likewise.
9677         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
9678         instance.
9679         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9680         * diagnostic-core.h (class auto_diagnostic_group): New class.
9681         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
9682         (diagnostic_report_diagnostic): Handle the first diagnostics within
9683         a group.
9684         (emit_diagnostic): Add auto_diagnostic_group instance.
9685         (inform): Likewise.
9686         (inform_n): Likewise.
9687         (warning): Likewise.
9688         (warning_at): Likewise.
9689         (warning_n): Likewise.
9690         (pedwarn): Likewise.
9691         (permerror): Likewise.
9692         (error): Likewise.
9693         (error_n): Likewise.
9694         (error_at): Likewise.
9695         (sorry): Likewise.
9696         (fatal_error): Likewise.
9697         (internal_error): Likewise.
9698         (internal_error_no_backtrace): Likewise.
9699         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9700         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9701         * diagnostic.h (struct diagnostic_context): Add fields
9702         "diagnostic_group_nesting_depth",
9703         "diagnostic_group_emission_count", "begin_group_cb",
9704         "end_group_cb".
9705         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9706         Add auto_diagnostic_group instance(s).
9707         (find_explicit_erroneous_behavior): Likewise.
9708         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9709         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9710         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9711         (gimplify_va_arg_expr): Likewise.
9712         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9713         (HSA_SORRY_AT): Likewise.
9714         * ipa-devirt.c (compare_virtual_tables): Likewise.
9715         (warn_odr): Likewise.
9716         * multiple_target.c (expand_target_clones): Likewise.
9717         * opts-common.c (cmdline_handle_error): Likewise.
9718         * reginfo.c (globalize_reg): Likewise.
9719         * substring-locations.c (format_warning_n_va): Likewise.
9720         * tree-inline.c (expand_call_inline): Likewise.
9721         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9722         * tree-ssa-loop-niter.c
9723         (do_warn_aggressive_loop_optimizations): Likewise.
9724         * tree-ssa-uninit.c (warn_uninit): Likewise.
9725         * tree.c (warn_deprecated_use): Likewise.
9727 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
9729         PR target/87014
9730         * config/i386/i386.md (eh_return): Always update EH return
9731         address in word_mode.
9733 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
9735         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9736         TARGET_SPLIT_COMPLEX_ARG is defined.
9738 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9740         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9742 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9744         PR target/86984
9745         * expr.c (expand_assignment): Assert that bitpos is positive.
9746         (store_field): Likewise
9747         (expand_expr_real_1): Make sure that bitpos is positive.
9748         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9749         integer overflow.
9751 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9753         * Makefile.in (CPP_ID_DATA_H): Delete.
9754         (CPP_INTERNAL_H): Don't add it.
9755         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9756         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9758 2018-08-20  Richard Biener  <rguenther@suse.de>
9760         PR tree-optimization/78655
9761         * tree-vrp.c (extract_range_from_binary_expr_1): Make
9762         pointer + offset nonnull if either operand is nonnull work.
9764 2018-08-20  Tom de Vries  <tdevries@suse.de>
9766         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9767         unless the referenced die describes the added property using
9768         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
9769         Otherwise, add a DW_AT_location to the referenced die.
9771 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
9773         PR target/86994
9774         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9775         register_operand when calling ix86_set_reg_reg_cost.
9776         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9777         Set *total to 0 for operands that satisfy x86_64_immediate_operand
9778         predicate and to 1 otherwise.
9780 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
9782         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9783         emit a diagnostic that it is not supported and reset the option.
9784         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9785         supported and consume the option.  (ASM_FINAL_SPEC): New.
9787 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9789         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9790         a sentence.
9792 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
9794         C-SKY port: Documentation
9796         * doc/extend.texi (C-SKY Function Attributes): New section.
9797         * doc/invoke.texi (Option Summary): Add C-SKY options.
9798         (C-SKY Options): New section.
9799         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9801 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9802             Huibin Wang  <huibin_wang@c-sky.com>
9803             Sandra Loosemore  <sandra@codesourcery.com>
9804             Chung-Lin Tang  <cltang@codesourcery.com>
9806         C-SKY port: Backend implementation
9808         * config/csky/*: New.
9809         * common/config/csky/*: New.
9811 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9812             Huibin Wang  <huibin_wang@c-sky.com>
9813             Sandra Loosemore  <sandra@codesourcery.com>
9814             Chung-Lin Tang  <cltang@codesourcery.com>
9815             Andrew Jenner  <andrew@codesourcery.com>
9817         C-SKY port: Configury
9819         * config.gcc (csky-*-*): New.
9820         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9821         * configure: Regenerated.
9823 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9825         * dump-context.h: Include "dumpfile.h".
9826         (dump_context::dump_printf_va): Convert final param from va_list
9827         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9828         ATTRIBUTE_GCC_DUMP_PRINTF.
9829         (dump_context::dump_printf_loc_va): Likewise.
9830         * dumpfile.c: Include "stringpool.h".
9831         (make_item_for_dump_printf_va): Delete.
9832         (make_item_for_dump_printf): Delete.
9833         (class dump_pretty_printer): New class.
9834         (dump_pretty_printer::dump_pretty_printer): New ctor.
9835         (dump_pretty_printer::emit_items): New member function.
9836         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9837         function.
9838         (dump_pretty_printer::emit_item): New member function.
9839         (dump_pretty_printer::stash_item): New member function.
9840         (dump_pretty_printer::format_decoder_cb): New member function.
9841         (dump_pretty_printer::decode_format): New member function.
9842         (dump_context::dump_printf_va): Reimplement in terms of
9843         dump_pretty_printer.
9844         (dump_context::dump_printf_loc_va): Convert final param from va_list
9845         to va_list *.
9846         (dump_context::begin_scope): Reimplement call to
9847         make_item_for_dump_printf.
9848         (dump_printf): Update for change to dump_printf_va.
9849         (dump_printf_loc): Likewise.
9850         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9851         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9852         with %T, %E, and %G.
9853         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9854         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9855         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9856         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9857         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9858         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9859         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9860         within a dump_printf_loc call to "%wu".
9861         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9862         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9863         missing space after "=".
9864         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9865         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9866         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9867         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9868         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9869         duplicate "vectorized" from message.
9871 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9873         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9874         polyNxK_t element's TYPE_STRING_FLAG.
9876 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9878         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9879         (they were unnamed before).  Fix comments.
9881 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9883         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9885 2018-08-17  Richard Biener  <rguenther@suse.de>
9887         PR tree-optimization/86841
9888         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9890 2018-08-17  Martin Liska  <mliska@suse.cz>
9892         * common.opt: Remove Warn, Init and Report for options with
9893         Ignore/Deprecated flag. Warning is done automatically for
9894         Deprecated flags.
9895         * config/i386/i386.opt: Likewise.
9896         * config/ia64/ia64.opt: Likewise.
9897         * config/rs6000/rs6000.opt: Likewise.
9898         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9899         Remove usage of flag_check_pointer_bounds.
9900         * lto-wrapper.c (merge_and_complain): Do not handle
9901         OPT_fcheck_pointer_bounds.
9902         (append_compiler_options): Likewise.
9903         * opt-functions.awk: Do not handle Deprecated.
9904         * optc-gen.awk: Check that Var, Report and Init are not
9905         used for an option with Ignore/Deprecated flag.
9906         * opts-common.c (decode_cmdline_option): Do not report
9907         CL_ERR_DEPRECATED.
9908         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9909         options.
9910         * opts.h (struct cl_option): Remove cl_deprecated flag.
9911         (CL_ERR_DEPRECATED): Remove error enum value.
9913 2018-08-17  Richard Biener  <rguenther@suse.de>
9915         PR middle-end/86505
9916         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9917         across a va-arg-pack using call adjust its return value accordingly.
9919 2018-08-16  Martin Sebor  <msebor@redhat.com>
9921         PR tree-optimization/86853
9922         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9923         (struct fmtresult): Add member and initialize it in ctors.
9924         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9925         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9926         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9927         (parse_directive): Handle %C same as %c.
9928         (sprintf_dom_walker::compute_format_length): Adjust.
9929         (is_call_safe): Adjust.
9931 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9933         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9934         for determining how to count the elements.
9935         * builtins.h (c_strlen): Adjust prototype.
9936         * expr.c (string_constant): Add new parameter mem_size.
9937         Set *mem_size appropriately.
9938         * expr.h (string_constant): Adjust protoype.
9939         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9940         * gimple-fold.h (get_range_strlen): Adjust prototype.
9941         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9942         (format_string): Call get_string_length with eltsize.
9944 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9946         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9947         to emit the span, rather than setting it as the prefix.
9949 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9951         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9952         "margin_char" parameter, defaulting to space.  Use it in place
9953         of pp_space for the initial part of the margin.
9954         (layout::print_leading_fixits): Use '+' when filling the margin
9955         of line-insertion fix-it hints.
9957 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9959         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9960         Delete.
9962 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9964         * config/rs6000/altivec.md: Don't set length attribute to the default
9965         value.
9966         * config/rs6000/darwin.md: Ditto.
9967         * config/rs6000/dfp.md: Ditto.
9968         * config/rs6000/htm.md: Ditto.
9969         * config/rs6000/rs6000.md: Ditto.
9970         * config/rs6000/sync.md: Ditto.
9971         * config/rs6000/vsx.md: Ditto.
9973 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9975         * config/rs6000/altivec.md: Don't set length attribute to the default
9976         value, for branch instructions.
9977         * config/rs6000/darwin.md: Ditto.
9978         * config/rs6000/rs6000.md: Ditto.
9980 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9982         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9983         (unnamed conditional branch define_insn): Set length to 4 or 8
9984         depending on offset.
9985         (<bd>_<mode>): Similar, for alternative 0.
9986         (<bd>tf_<mode>): Ditto.
9988 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9990         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9992 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9994         * doc/rtl.texi: Replace old RTX class names with new names.
9997 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9999         * expmed.h (canonicalize_comparison): New declaration.
10000         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
10001         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
10002         * optabs.c (prepare_cmp_insn): Likewise.
10003         * rtl.h (unsigned_condition_p): New function which checks if a
10004         comparison operator is unsigned.
10006 2018-08-16  Nathan Sidwell  <nathan@acm.org>
10008         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
10009         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
10011 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
10013         PR target/84711
10014         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
10015         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
10016         (mov<mov>): ..this and enable unconditionally.
10018 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
10020         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
10022 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
10024         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
10025         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
10026         "Common" with "Target".
10028 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
10030         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
10031         * doc/invoke.texi (mmitigate-rop): Remove.
10032         * config/i386/i386.c: Do not include "regrename.h".
10033         (ix86_rop_should_change_byte_p, reg_encoded_number)
10034         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
10035         Remove.
10036         (ix86_reorg): Remove call to ix86_mitigate_rop.
10037         * config/i386/i386.md (attr "modrm_class"): Remove.
10038         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
10039         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
10040         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
10042 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
10044         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
10045         allow folding of mergeh() and mergel() for the float and double types.
10046         (fold_mergehl_helper): Rework to handle building a permute tree
10047         for float vectors.
10049 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
10051         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
10052         for TARGET_SSE.
10054 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
10056         * common.opt (fdiagnostics-show-labels): New option.
10057         * diagnostic-show-locus.c (class layout_range): Add field
10058         "m_label".
10059         (class layout): Add field "m_show_labels_p".
10060         (layout_range::layout_range): Add param "label" and use it to
10061         initialize m_label.
10062         (make_range): Pass in NULL for new "label" param of layout_range's
10063         ctor.
10064         (layout::layout): Initialize m_show_labels_p.
10065         (layout::maybe_add_location_range): Pass in loc_range->m_label
10066         when constructing layout_range instances.
10067         (struct line_label): New struct.
10068         (layout::print_any_labels): New member function.
10069         (layout::print_line): Call it if label-printing is enabled.
10070         (selftest::test_one_liner_labels): New test.
10071         (selftest::test_diagnostic_show_locus_one_liner): Call it.
10072         * diagnostic.c (diagnostic_initialize): Initialize
10073         context->show_labels_p.
10074         * diagnostic.h (struct diagnostic_context): Add field
10075         "show_labels_p".
10076         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
10077         -fno-diagnostics-show-labels.
10078         * dwarf2out.c (gen_producer_string): Add
10079         OPT_fdiagnostics_show_labels to the ignored options.
10080         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
10081         param.
10082         (gcc_rich_location::maybe_add_expr): Likewise.
10083         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
10084         label" param, defaulting to NULL.
10085         (gcc_rich_location::add_expr): Add "label" param.
10086         (gcc_rich_location::maybe_add_expr): Likewise.
10087         (class text_range_label): New class.
10088         (class range_label_for_type_mismatch): New class.
10089         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
10090         of format_warning_va.
10091         (fmtwarn_n): Likewise for new params of format_warning_n_va.
10092         * lto-wrapper.c (merge_and_complain): Add
10093         OPT_fdiagnostics_show_labels to the "pick one setting" options.
10094         (append_compiler_options): Likewise to the dropped options.
10095         (append_diag_options): Likewise to the passed-on options.
10096         * opts.c (common_handle_option): Handle the new option.
10097         * selftest-diagnostic.c
10098         (test_diagnostic_context::test_diagnostic_context): Enable
10099         show_labels_p.
10100         * substring-locations.c: Include "gcc-rich-location.h".
10101         (format_warning_n_va): Add "fmt_label" and "param_label" params
10102         and use them as appropriate.
10103         (format_warning_va): Add "fmt_label" and "param_label" params,
10104         passing them on to format_warning_n_va.
10105         (format_warning_at_substring): Likewise.
10106         (format_warning_at_substring_n): Likewise.
10107         * substring-locations.h (format_warning_va): Add "fmt_label" and
10108         "param_label" params.
10109         (format_warning_n_va): Likewise.
10110         (format_warning_at_substring): Likewise.
10111         (format_warning_at_substring_n): Likewise.
10112         * toplev.c (general_init): Initialize global_dc->show_labels_p.
10114 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
10116         PR testsuite/86519
10117         * builtins.c (expand_builtin_memcmp): Do not expand the call
10118         when overflow is detected.
10120 2018-08-15  Martin Sebor  <msebor@redhat.com>
10122         PR tree-optimization/71625
10123         * config/aarch64/aarch64-builtins.c
10124         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
10126 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
10128         * config/s390/s390.c (s390_reorg): Remove loop.
10130 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
10132         * config/darwin.c
10133          (darwin_function_switched_text_sections): Delete.
10134         * gcc/config/darwin.h
10135          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
10137 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
10139         PR target/81685
10140         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
10141         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
10142         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
10144 2018-08-15  Martin Liska  <mliska@suse.cz>
10146         PR tree-optimization/86925
10147         * predict.c (expr_expected_value_1): When taking
10148         later predictor, assign also probability.
10149         Use fold_build2_initializer_loc in order to fold
10150         the expression in -frounding-math.
10152 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
10154         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
10155         patterns.
10156         (expand_vec_perm_1): Try the new method.
10158 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
10160         PR target/86547
10161         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
10162         Check whether lra_live_max_point is 0 before dividing.
10164 2018-08-14  Martin Sebor  <msebor@redhat.com>
10166         PR tree-optimization/86650
10167         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
10168         (vrp_prop::check_mem_ref): Same.
10170 2018-08-13  Liu Hao <lh_mouse@126.com>
10172         * pretty-print.c (eat_esc_sequence): Swap the foreground and
10173         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
10174         and clear it thereafter, as it only works for DBCS.
10176 2018-08-13  Liu Hao <lh_mouse@126.com>
10178         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
10179         handle returned by _get_osf_handle().
10181 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
10183         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
10184         for folding vec_perm.
10186 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
10188         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
10189         Add support for gimple-folding of vec_pack() and vec_unpack()
10190         intrinsics.
10192 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
10194         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
10195         vec_xst variants to the list.
10196         (rs6000_gimple_fold_builtin): Add support for folding unaligned
10197         vector loads and stores.
10199 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
10201         * config.gcc (rs6000-ibm-aix4.x): Delete.
10202         (rs6000-ibm-aix5.1): Delete.
10203         (rs6000-ibm-aix5.2): Delete.
10204         (rs6000-ibm-aix5.3): Delete.
10205         * config/rs6000/aix43.h: Delete.
10206         * config/rs6000/aix51.h: Delete.
10207         * config/rs6000/aix52.h: Delete.
10208         * config/rs6000/t-aix43: Delete.
10210 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
10212         * config/s390/s390.c (s390_decompose_constant_pool_ref):
10213         New function.
10214         (s390_decompose_address): Factor out constant pool ref
10215         decomposition.
10217 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10219         * config/nds32/nds32-predicates.c
10220         (nds32_can_use_bclr_p): Change return type as bool.
10221         (nds32_can_use_bset_p): Ditto.
10222         (nds32_can_use_btgl_p): Ditto.
10223         (nds32_can_use_bitci_p): Ditto.
10224         * config/nds32/nds32-protos.h
10225         (nds32_can_use_bclr_p): Change declaration.
10226         (nds32_can_use_bset_p): Ditto.
10227         (nds32_can_use_btgl_p): Ditto.
10228         (nds32_can_use_bitci_p): Ditto.
10230 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10232         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
10233         Support -msched-prolog-epilog option.
10234         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
10236 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10238         * common/config/nds32/nds32-common.c
10239         (nds32_option_optimization_table): Enalbe -malways-align.
10241 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10243         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
10244         extra_headers.
10245         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
10246         OPT_misr_secure_ case.
10247         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
10248         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
10249         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
10250         secure attribute.
10251         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
10252         (nds32_isr_info): New field security_level.
10253         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
10254         * config/nds32/nds32.md (return_internal): Consider critical attribute.
10255         * config/nds32/nds32.opt (misr-secure): New option.
10256         * config/nds32/nds32_init.inc: New file.
10257         * config/nds32/nds32_isr.h: New file.
10259 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
10261         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
10262         Update comment for atomic instructions.
10263         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
10264         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
10265         Remove.
10266         (atomic_loaddi): Revise fence expansion to only emit fence prior to
10267         load for __ATOMIC_SEQ_CST model.
10268         (atomic_loaddi_1): Remove float register target.
10269         (atomic_storedi): Handle CONST_INT values.
10270         (atomic_storedi_1): Remove float register source.  Add special case
10271         for zero value.
10272         (memory_barrier): New expander and insn.
10274 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
10276         PR tree-optimization/86835
10277         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
10278         new_stmt after def_gsi, make sure to insert new_square_stmt after
10279         that stmt, not 2 stmts before it.
10281 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
10283         PR target/82418
10284         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
10285         instead of SWI48.
10287 2018-08-10  Martin Liska  <mliska@suse.cz>
10289         PR target/83610
10290         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
10291         function type.
10292         * builtins.c (expand_builtin_expect_with_probability):
10293         New function.
10294         (expand_builtin_expect_with_probability): New function.
10295         (build_builtin_expect_predicate): Add new argumnet probability
10296         for BUILT_IN_EXPECT_WITH_PROBABILITY.
10297         (fold_builtin_expect):
10298         (fold_builtin_2):
10299         (fold_builtin_3):
10300         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
10301         * builtins.h (fold_builtin_expect): Set new argument.
10302         * doc/extend.texi: Document __builtin_expect_with_probability.
10303         * doc/invoke.texi: Likewise.
10304         * gimple-fold.c (gimple_fold_call): Pass new argument.
10305         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
10306         also BUILT_IN_EXPECT_WITH_PROBABILITY.
10307         * predict.c (get_predictor_value): New function.
10308         (expr_expected_value): Add new argument probability. Assume
10309         that predictor and probability are always non-null.
10310         (expr_expected_value_1): Likewise.  For __builtin_expect and
10311         __builtin_expect_with_probability set probability.  Handle
10312         combination in binary expressions.
10313         (tree_predict_by_opcode): Simplify code by simply calling
10314         get_predictor_value.
10315         (pass_strip_predict_hints::execute): Add handling of
10316         BUILT_IN_EXPECT_WITH_PROBABILITY.
10317         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
10318         new predictor.
10319         * tree.h (DECL_BUILT_IN_P): New function.
10321 2018-08-10  Martin Liska  <mliska@suse.cz>
10323         PR tree-optimization/85799
10324         * passes.def: Add argument for pass_strip_predict_hints.
10325         * predict.c (class pass_strip_predict_hints): Add new argument
10326         early_p.
10327         (strip_predictor_early): New function.
10328         (pass_strip_predict_hints::execute): Call the function to
10329         strip predictors.
10330         (strip_predict_hints): New function.
10331         * predict.def: Fix comment.
10333 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
10335         * Makefile.in: Clarify which tm.texi to copy over to assert the
10336         right to grant a GFDL license for all.
10338 2018-08-09  Jeff Law  <law@redhat.com>
10340         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
10341         unused variable.
10343 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
10345         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
10346         prototype.
10348 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10350         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
10351         reductions for variable-length vectors.
10353 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
10355         PR other/84889
10356         * common.opt (fdiagnostics-show-line-numbers): New option.
10357         * diagnostic-show-locus.c (class layout): Add fields
10358         "m_show_line_numbers_p" and "m_linenum_width";
10359         (num_digits): New function.
10360         (test_num_digits): New function.
10361         (layout::layout): Initialize new fields.  Update m_x_offset
10362         logic to handle any left margin.
10363         (layout::print_source_line): Print line number when requested.
10364         (layout::start_annotation_line): New member function.
10365         (layout::print_annotation_line): Call it.
10366         (layout::print_leading_fixits): Likewise.
10367         (layout::print_trailing_fixits): Likewise.  Update calls to
10368         move_to_column for new parameter.
10369         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
10370         it to potentially call start_annotation_line.
10371         (layout::show_ruler): Call start_annotation_line.
10372         (selftest::test_line_numbers_multiline_range): New selftest.
10373         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
10374         and selftest::test_line_numbers_multiline_range.
10375         * diagnostic.c (diagnostic_initialize): Initialize
10376         show_line_numbers_p.
10377         * diagnostic.h (struct diagnostic_context): Add field
10378         "show_line_numbers_p".
10379         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
10380         -fno-diagnostics-show-line-numbers.
10381         * dwarf2out.c (gen_producer_string): Add
10382         OPT_fdiagnostics_show_line_numbers to the ignored options.
10383         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
10384         one setting" options.
10385         (append_compiler_options): Likewise to the dropped options.
10386         (append_diag_options): Likewise to the passed-on options.
10387         * opts.c (common_handle_option): Handle the new option.
10388         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
10390 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10392         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
10393         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
10394         third argument of type "const signed char" to descriptions of
10395         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
10396         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
10397         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
10398         __builtin_bcdsub_ov functions.
10400 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10402         PR tree-optimization/86858
10403         * tree-vect-loop.c (vect_is_simple_reduction): Restore
10404         flow_bb_inside_loop_p calls.
10406 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10408         PR tree-optimization/86871
10409         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
10410         instead of gimple_assign_lhs.
10412 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
10414         PR target/86887
10415         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
10416         register constraint to operand 0.
10417         (add<mode>3_carryinC): Likewise.
10418         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
10420 2018-08-09  Martin Liska  <mliska@suse.cz>
10422         PR c/86895
10423         * common.opt: Remove extra line.
10425 2018-08-09  Martin Liska  <mliska@suse.cz>
10427         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
10428         at the end of a line, make first letter capital and end up
10429         a sentence with a dot.
10430         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
10431         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
10432         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
10433         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
10434         (PARAM_MAX_ISL_OPERATIONS): Likewise.
10435         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
10436         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
10437         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
10438         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
10439         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
10440         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
10441         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
10442         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
10443         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
10444         (PARAM_TREE_REASSOC_WIDTH): Likewise.
10445         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
10446         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
10447         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
10449 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
10451         PR target/84332
10452         * config/s390/s390.c (s390_option_override_internal): Reduce the
10453         stack-clash-protection-probe-interval param if it would be too big
10454         for z900.
10456 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
10458         PR target/46179
10459         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
10460         * config/m68k/m68k.c (handle_move_double): Don't call
10461         m68k_final_prescan_insn.
10462         (m68k_adjust_decorated_operand): Renamed from
10463         m68k_final_prescan_insn, remove first and third operand and
10464         simplify.
10465         (print_operand): Call it.
10466         (print_operand_address): Call it.
10468 2018-08-08  Nathan Sidwell  <nathan@acm.org>
10470         * diagnostic.c (diagnostic_report_current_module): Use
10471         linemap_included_from & linemap_included_from_linemap.
10473 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
10475         * config/aarch64/aarch64-cores.def: Add phecda core.
10476         * config/aarch64/aarch64-tune.md: Regenerate.
10477         * doc/invoke.texi: Add phecda core.
10479 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
10481         PR target/85295
10482         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
10483         definitions.
10484         * config/s390/s390.md ("movti"): Add more alternatives for
10485         constant to GPR copies.
10487 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
10489         * config/s390/s390.c: Fix whitespace damage throughout the file.
10490         * config/s390/s390.h: Likewise.
10491         * config/s390/tpf.h: Likewise.
10493 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
10495         * config/s390/s390.c (s390_loadrelative_operand_p):
10496         Remove TARGET_CPU_ZARCH usages.
10497         (s390_rtx_costs): Likewise.
10498         (s390_legitimate_constant_p): Likewise.
10499         (s390_cannot_force_const_mem): Likewise.
10500         (legitimate_reload_constant_p): Likewise.
10501         (s390_preferred_reload_class): Likewise.
10502         (legitimize_pic_address): Likewise.
10503         (legitimize_tls_address): Likewise.
10504         (s390_split_branches): Removed.
10505         (s390_add_execute): Removed.
10506         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
10507         (s390_mainpool_start): Likewise.
10508         (s390_mainpool_finish): Likewise.
10509         (s390_mainpool_cancel): Removed.
10510         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
10511         (s390_chunkify_cancel): Likewise.
10512         (s390_return_addr_rtx): Likewise.
10513         (s390_register_info): Remove split_branches_pending_p uages.
10514         (s390_optimize_register_info): Likewise.
10515         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
10516         split_branches_pending_p usages.
10517         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
10518         (s390_load_got): Likewise.
10519         (s390_expand_split_stack_prologue): Likewise.
10520         (output_asm_nops): Likewise.
10521         (s390_function_profiler): Likewise.
10522         (s390_emit_call): Likewise.
10523         (s390_conditional_register_usage): Likewise.
10524         (s390_optimize_prologue): Likewise.
10525         (s390_reorg): Remove TARGET_CPU_ZARCH and
10526         split_branches_pending_p usages.
10527         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
10528         usages.
10529         (s390_output_indirect_thunk_function): Likewise.
10530         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
10531         (TARGET_CPU_ZARCH_P): Removed.
10532         (struct machine_function): Remove split_branches_pending_p.
10533         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
10535 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
10537         * common/config/s390/s390-common.c (processor_flags_table):
10538         Remove flags.
10539         * config.gcc: Remove with_arch/with_tune support.
10540         * config/s390/2064.md: Remove cpu attribute comparisons.
10541         * config/s390/driver-native.c (s390_host_detect_local_cpu):
10542         Remove MTN.
10543         * config/s390/linux.h (ASM_SPEC):
10544         Remove -march support.
10545         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
10546         Use a table to get an arch level.
10547         * config/s390/s390-opts.h (enum processor_type):
10548         Remove enum values.
10549         * config/s390/s390.c
10550         (processor_table): Remove entries, add arch_level values.
10551         (s390_issue_rate): Remove cases.
10552         (s390_option_override): Adjust
10553         s390_option_override_internal() call.
10554         (s390_option_override_internal): Remove deprecation warning.
10555         (s390_valid_target_attribute_tree): Adjust
10556         s390_option_override_internal() call.
10557         * config/s390/s390.h (struct s390_processor):
10558         Share with s390-c.c, add arch_level field.
10559         * config/s390/s390.md:
10560         Remove occurrences in cpu attribute.
10561         * config/s390/s390.opt: Remove -march/-mtune support.
10562         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
10563         * doc/invoke.texi: Remove deprecation warning.
10565 2018-08-08  Luis Machado  <luis.machado@linaro.org>
10567         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
10568         global.
10569         (qdf24xx_tunings): Set vector cost structure to
10570         qdf24xx_vector_cost.
10572         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
10573         <register_sextend>: Set to 3.
10575 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
10577         PR target/86838
10578         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
10579         * config/aarch64/aarch64-simd.md
10580         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
10581         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
10582         (aarch64_frecpx<mode>): ...this new pattern.
10583         * config/aarch64/aarch64-simd-builtins.def: Remove comment
10584         about aarch64_frecp<FRECP:frecp_suffix><mode>.
10586 2018-08-07  Martin Liska  <mliska@suse.cz>
10588         PR middle-end/83023
10589         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
10590         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
10591         * predict.def (PRED_MALLOC_NONNULL): New predictor.
10592         * doc/extend.texi: Document that malloc attribute adds
10593         hit to compiler.
10595 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
10597         PR target/86785
10598         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10599         Define to speculation_safe_value_not_needed.
10601 2018-08-06  Jeff Law  <law@redhat.com>
10603         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
10604         the vr_values instance to cprop_into_stmt.
10605         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
10606         (cprop_operand): Also query EVRP to determine if OP is a constant.
10608 2018-08-06  Nathan Sidwell  <nathan@acm.org>
10610         * diagnostic.c (diagnostic_report_current_module): Reroll
10611         included-at loop.  Translate text.
10613 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
10615         * function-tests.c (selftest::test_expansion_to_rtl): Call
10616         free_after_compilation.
10618 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10620         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
10622 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
10624         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
10625         loops with memory block operations from getting unrolled.
10627 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
10629         PR target/86807
10630         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10631         Define to speculation_safe_value_not_needed.
10633 2018-08-06  Jeff Law  <law@redhat.com>
10635         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
10636         assert.
10638 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10640         PR target/86662
10641         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
10642         with all enabled __intN types.
10644         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
10646 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10648         * alias.c (record_set): Check for clobber high.
10649         * cfgexpand.c (expand_gimple_stmt): Likewise.
10650         * combine-stack-adj.c (single_set_for_csa): Likewise.
10651         * combine.c (find_single_use_1): Likewise.
10652         (set_nonzero_bits_and_sign_copies): Likewise.
10653         (get_combine_src_dest): Likewise.
10654         (is_parallel_of_n_reg_sets): Likewise.
10655         (try_combine): Likewise.
10656         (record_dead_and_set_regs_1): Likewise.
10657         (reg_dead_at_p_1): Likewise.
10658         (reg_dead_at_p): Likewise.
10659         * dce.c (deletable_insn_p): Likewise.
10660         (mark_nonreg_stores_1): Likewise.
10661         (mark_nonreg_stores_2): Likewise.
10662         * df-scan.c (df_find_hard_reg_defs): Likewise.
10663         (df_uses_record): Likewise.
10664         (df_get_call_refs): Likewise.
10665         * dwarf2out.c (mem_loc_descriptor): Likewise.
10666         * haifa-sched.c (haifa_classify_rtx): Likewise.
10667         * ira-build.c (create_insn_allocnos): Likewise.
10668         * ira-costs.c (scan_one_insn): Likewise.
10669         * ira.c (equiv_init_movable_p): Likewise.
10670         (rtx_moveable_p): Likewise.
10671         (interesting_dest_for_shprep): Likewise.
10672         * jump.c (mark_jump_label_1): Likewise.
10673         * postreload-gcse.c (record_opr_changes): Likewise.
10674         * postreload.c (reload_cse_simplify): Likewise.
10675         (struct reg_use): Add source expr.
10676         (reload_combine): Check for clobber high.
10677         (reload_combine_note_use): Likewise.
10678         (reload_cse_move2add): Likewise.
10679         (move2add_note_store): Likewise.
10680         * print-rtl.c (print_pattern): Likewise.
10681         * recog.c (decode_asm_operands): Likewise.
10682         (store_data_bypass_p): Likewise.
10683         (if_test_bypass_p): Likewise.
10684         * regcprop.c (kill_clobbered_value): Likewise.
10685         (kill_set_value): Likewise.
10686         * reginfo.c (reg_scan_mark_refs): Likewise.
10687         * reload1.c (maybe_fix_stack_asms): Likewise.
10688         (eliminate_regs_1): Likewise.
10689         (elimination_effects): Likewise.
10690         (mark_not_eliminable): Likewise.
10691         (scan_paradoxical_subregs): Likewise.
10692         (forget_old_reloads_1): Likewise.
10693         * reorg.c (find_end_label): Likewise.
10694         (try_merge_delay_insns): Likewise.
10695         (redundant_insn): Likewise.
10696         (own_thread_p): Likewise.
10697         (fill_simple_delay_slots): Likewise.
10698         (fill_slots_from_thread): Likewise.
10699         (dbr_schedule): Likewise.
10700         * resource.c (update_live_status): Likewise.
10701         (mark_referenced_resources): Likewise.
10702         (mark_set_resources): Likewise.
10703         * rtl.c (copy_rtx): Likewise.
10704         * rtlanal.c (reg_referenced_p): Likewise.
10705         (single_set_2): Likewise.
10706         (noop_move_p): Likewise.
10707         (note_stores): Likewise.
10708         * sched-deps.c (sched_analyze_reg): Likewise.
10709         (sched_analyze_insn): Likewise.
10711 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10713         * cse.c (invalidate_reg): New function extracted from...
10714         (invalidate): ...here.
10715         (canonicalize_insn): Check for clobber high.
10716         (invalidate_from_clobbers): invalidate clobber highs.
10717         (invalidate_from_sets_and_clobbers): Likewise.
10718         (count_reg_usage): Check for clobber high.
10719         (insn_live_p): Likewise.
10720         * cselib.c (cselib_expand_value_rtx_1):Likewise.
10721         (cselib_invalidate_regno): Check for clobber in setter.
10722         (cselib_invalidate_rtx): Pass through setter.
10723         (cselib_invalidate_rtx_note_stores):
10724         (cselib_process_insn): Check for clobber high.
10725         * cselib.h (cselib_invalidate_rtx): Add operand.
10727 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10729         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10730         (mark_not_eliminable): Likewise.
10731         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10732         * lra-lives.c (process_bb_lives): Check for clobber high.
10733         * lra.c (new_insn_reg): Remember clobber highs.
10734         (collect_non_operand_hard_regs): Check for clobber high.
10735         (lra_set_insn_recog_data): Likewise.
10736         (add_regs_to_insn_regno_info): Likewise.
10737         (lra_update_insn_regno_info): Likewise.
10739 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10741         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10742         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10744 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10746         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10747         (copy_insn_1): Likewise.
10748         (gen_hard_reg_clobber_high): New gen function.
10749         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10750         * genemit.c (gen_exp): Likewise.
10751         (gen_emit_seq): Pass through info.
10752         (gen_insn): Check for CLOBBER_HIGH.
10753         (gen_expand): Pass through info.
10754         (gen_split): Likewise.
10755         (output_add_clobbers): Likewise.
10756         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10757         (remove_clobbers): Likewise.
10758         * rtl.h (gen_hard_reg_clobber_high): New declaration.
10760 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10762         * doc/rtl.texi (clobber_high): Add.
10763         (parallel): Add in clobber high
10764         * rtl.c (rtl_check_failed_code3): Add function.
10765         * rtl.def (CLOBBER_HIGH): Add expression.
10766         * rtl.h (RTL_CHECKC3): Add macro.
10767         (rtl_check_failed_code3): Add declaration.
10768         (XC3EXP): Add macro.
10770 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
10772         PR target/86386
10773         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10774         cfun->machine->max_used_stack_alignment if needed.
10776 2018-08-04  Martin Sebor  <msebor@redhat.com>
10778         PR tree-optimization/86571
10779         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10780         NaN output to 4.
10782 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
10784         PR target/86799
10785         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10786         Define.
10788 2018-08-03  Jeff Law  <law@redhat.com>
10790         PR target/86795
10791         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10792         Define to speculation_safe_value_not_needed.
10794 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
10796         * doc/gcov.texi (-x): Remove duplicate "to".
10797         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10798         (-Wif-not-aligned): Remove duplicate "is".
10799         (-flto): Remove duplicate "the".
10800         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10801         duplicate "v5.00.b".
10802         (MSP430 Options): Remove duplicate "and" from the description
10803         of "-mgprel-sec=regexp".
10804         (x86 Options): Remove duplicate copies of "vmldLog102" and
10805         vmlsLog104 from description of "-mveclibabi=type".
10807 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
10809         * internal-fn.h (first_commutative_argument): Declare.
10810         * internal-fn.c (first_commutative_argument): New function.
10811         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10812         restrictions for pattern statements.  Use first_commutative_argument
10813         to look for commutative operands in calls to internal functions.
10815 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10817         * Makefile.in (wide-int-range.o): New.
10818         * tree-vrp.c: Move all the wide_int_* functions to...
10819         * wide-int-range.cc: ...here.
10820         * tree-vrp.h: Move all the wide_int_* prototypes to...
10821         * wide-int-range.h: ...here.
10823 2018-08-03  Tom de Vries  <tdevries@suse.de>
10825         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10826         UI_NONE.
10827         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10828         * except.c (output_function_exception_table): Do early exit if
10829         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10831 2018-08-03  Martin Liska  <mliska@suse.cz>
10833         * predict.c (dump_prediction): Change to 2 digits
10834         in fraction part.
10836 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10838         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10839         neon_dup_q to...
10840         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10842 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10844         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10845         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10846         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10847         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10848         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10850 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10852         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10853         pp_take_prefix when saving the existing prefix.
10854         * diagnostic.c (diagnostic_append_note): Likewise.
10855         * langhooks.c (lhd_print_error_function): Likewise.
10856         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10857         param's type.  Free the existing prefix.
10858         (pp_take_prefix): New function.
10859         (pretty_printer::pretty_printer): Drop the prefix parameter.
10860         Rename the length parameter to match the comment.
10861         (pretty_printer::~pretty_printer): Free the prefix.
10862         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10863         parameter.
10864         (struct pretty_printer): Drop the "const" from "prefix" field's
10865         type and clarify memory management.
10866         (pp_set_prefix): Drop the "const" from the 2nd param.
10867         (pp_take_prefix): New decl.
10869 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10871         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10872         for word_mode_ok here instead of passing as argument.
10873         (expand_block_compare): Change select_block_compare_mode() call.
10874         (expand_strncmp_gpr_sequence): New function.
10875         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10877 2018-08-02  Jeff Law  <law@redhat.com>
10879         PR target/86790
10880         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10881         Define to speculation_safe_value_not_needed.
10883         PR target/86784
10884         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10885         Define to speculation_safe_value_not_needed.
10887 2018-08-02  Tom de Vries  <tdevries@suse.de>
10889         PR target/86660
10890         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10891         function.  Return UI_TARGET unconditionally.
10892         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10893         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10895 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10897         * genemit.c (print_overload_test): Fix typo.
10899 2018-08-02  Richard Biener  <rguenther@suse.de>
10901         PR tree-optimization/86816
10902         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10903         which checks for value availability before querying it.
10904         (gvn_uses_equal): Use it.
10905         (same_succ_hash): Likewise.
10906         (gimple_equal_p): Likewise.
10908 2018-08-02  Nick Clifton  <nickc@redhat.com>
10910         PR target/86813
10911         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10912         Define to speculation_safe_value_not_needed.
10914         PR target/86810
10915         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10916         Define to speculation_safe_value_not_needed.
10918         PR target/86810
10919         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10920         Define to speculation_safe_value_not_needed.
10922         PR target/86803
10923         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10924         Define to speculation_safe_value_not_needed.
10926         PR target/86797
10927         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10928         Define to speculation_safe_value_not_needed.
10930         PR target/86791
10931         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10932         Define to speculation_safe_value_not_needed.
10934         PR target/86789
10935         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10936         Define to speculation_safe_value_not_needed.
10938         PR target/86787
10939         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10940         Define to speculation_safe_value_not_needed.
10942         PR target/86782
10943         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10944         speculation_safe_value_not_needed.
10946         PR target/86781
10947         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10948         to speculation_safe_value_not_needed.
10950 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10952         * doc/md.texi: Expand the documentation of instruction names
10953         to mention port-local uses.  Document '@' in pattern names.
10954         * read-md.h (overloaded_instance, overloaded_name): New structs.
10955         (mapping): Declare.
10956         (md_reader::handle_overloaded_name): New member function.
10957         (md_reader::get_overloads): Likewise.
10958         (md_reader::m_first_overload): New member variable.
10959         (md_reader::m_next_overload_ptr): Likewise.
10960         (md_reader::m_overloads_htab): Likewise.
10961         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10962         m_next_overload_ptr and m_overloads_htab.
10963         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10964         (get_mode_token, get_code_token, get_int_token): New functions.
10965         (map_attr_string): Add an optional argument that passes back
10966         the associated iterator.
10967         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10968         (md_reader::handle_overloaded_name, add_overload_instance): New
10969         functions.
10970         (apply_iterators): Handle '@' names.  Report an error if '@'
10971         is used without iterators.
10972         (initialize_iterators): Initialize the new iterator_group fields.
10973         * genopinit.c (handle_overloaded_code_for)
10974         (handle_overloaded_gen): New functions.
10975         (main): Use them to print declarations of maybe_code_for_* and
10976         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10977         * genemit.c (print_overload_arguments, print_overload_test)
10978         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10979         (main): Use it to print definitions of maybe_code_for_* and
10980         maybe_gen_* functions.
10981         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10982         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10983         instead of explicit mode checks.
10984         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10985         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10986         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10987         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10988         (aarch64_expand_compare_and_swap): Likewise
10989         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10990         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10991         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10992         (aarch64_constant_pool_reload_icode): Delete.
10993         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10994         instead of aarch64_constant_pool_reload_icode.  Use
10995         code_for_aarch64_reload_mov instead of explicit mode checks.
10996         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10997         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10998         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10999         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
11000         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
11001         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
11002         (aarch64_atomic_load_op_code): Delete.
11003         (aarch64_emit_atomic_load_op): Likewise.
11004         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
11005         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
11006         instead of aarch64_emit_atomic_load_op.
11007         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
11008         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
11009         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
11010         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
11011         character before the pattern name.
11012         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
11013         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
11014         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
11015         (aarch64_frecps<mode>): Likewise.
11016         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
11017         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
11018         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
11019         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
11020         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
11022 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
11024         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
11025         Allow HFmode constants if TARGET_FP_F16INST.
11027 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
11029         PR target/86014
11030         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
11031         No longer check last store for clobber of address register.
11033 2018-08-02  Martin Liska  <mliska@suse.cz>
11035         PR gcov-profile/86817
11036         * gcov.c (process_all_functions): New function.
11037         (main): Call it.
11038         (process_file): Move functions processing to
11039         process_all_functions.
11041 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
11043         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
11044         "const" to the "gimple *" and "rtx_insn *" parameters.
11045         * dumpfile.h (dump_user_location_t::dump_user_location_t):
11046         Likewise.
11047         (dump_location_t::dump_location_t): Likewise.
11049 2018-08-01  Martin Sebor  <msebor@redhat.com>
11051         PR tree-optimization/86650
11052         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
11053         rather than a "gcall *".  Directly pass the data of interest
11054         to percent_K_format, rather than building a temporary CALL_EXPR
11055         to hold it.
11056         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
11057         (gimple_fold_builtin_strncat): Adjust.
11058         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
11059         gcall* argument with gimple*.
11060         * gimple-ssa-warn-restrict.c (check_call): Same.
11061         (wrestrict_dom_walker::before_dom_children): Same.
11062         (builtin_access::builtin_access): Same.
11063         (check_bounds_or_overlap): Same
11064         (maybe_diag_overlap): Same.
11065         (maybe_diag_offset_bounds): Same.
11066         * tree-diagnostic.c (default_tree_printer): Move usage of
11067         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
11068         to this callsite.
11069         * tree-pretty-print.c (percent_K_format): Add argument.
11070         * tree-pretty-print.h: Add argument.
11071         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
11072         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
11073         (maybe_diag_stxncpy_trunc): Same.
11074         (handle_builtin_stxncpy): Same.
11075         (handle_builtin_strcat): Same.
11077 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11079         * match.pd: Optimise pointer range checks.
11081 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11083         PR tree-optimization/86758
11084         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
11085         to remove pattern statements.
11087 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11089         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
11090         result of dfs_enumerate_from when constructing stmt_vec_infos,
11091         instead of additionally calling get_loop_body.
11093 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11095         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
11096         parameter.
11097         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
11098         When creating an iv, assert that the step is not known to be zero.
11099         (vect_setup_realignment): Update call accordingly.
11100         * tree-vect-stmts.c (vectorizable_store): Likewise.
11101         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
11103 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11105         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
11106         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
11107         (vectorizable_reduction): Likewise.
11108         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11109         (vect_detect_hybrid_slp_stmts): Likewise.
11110         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11112 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
11114         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
11115         (wide_int_set_zero_nonzero_bits): ...this.
11116         (zero_nonzero_bits_from_vr): Rename to...
11117         (vrp_set_zero_nonzero_bits): ...this.
11118         (extract_range_from_multiplicative_op_1): Abstract wide int
11119         code...
11120         (wide_int_range_multiplicative_op): ...here.
11121         (extract_range_from_binary_expr_1): Extract wide int binary
11122         operations into their own functions.
11123         (wide_int_range_lshift): New.
11124         (wide_int_range_can_optimize_bit_op): New.
11125         (wide_int_range_shift_undefined_p): New.
11126         (wide_int_range_bit_xor): New.
11127         (wide_int_range_bit_ior): New.
11128         (wide_int_range_bit_and): New.
11129         (wide_int_range_trunc_mod): New.
11130         (extract_range_into_wide_ints): New.
11131         (vrp_shift_undefined_p): New.
11132         (extract_range_from_multiplicative_op): New.
11133         (vrp_can_optimize_bit_op): New.
11134         * tree-vrp.h (value_range::dump): New.
11135         (wide_int_range_multiplicative_op): New.
11136         (wide_int_range_lshift):New.
11137         (wide_int_range_shift_undefined_p): New.
11138         (wide_int_range_bit_xor): New.
11139         (wide_int_range_bit_ior): New.
11140         (wide_int_range_bit_and): New.
11141         (wide_int_range_trunc_mod): New.
11142         (zero_nonzero_bits_from_bounds): Rename to...
11143         (wide_int_set_zero_nonzero_bits): ...this.
11144         (zero_nonzero_bits_from_vr): Rename to...
11145         (vrp_set_zero_nonzero_bits): ...this.
11146         (range_easy_mask_min_max): Rename to...
11147         (wide_int_range_can_optimize_bit_op): this.
11148         * vr-values.c (simplify_bit_ops_using_ranges): Rename
11149         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
11151 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11153         * tree-vectorizer.h (vect_orig_stmt): New function.
11154         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
11155         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
11156         (vect_create_epilog_for_reduction): Likewise.
11157         (vectorizable_live_operation): Likewise.
11158         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11159         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
11160         * tree-vect-stmts.c (vectorizable_call): Likewise.
11161         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
11163 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11165         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
11166         argument.
11167         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
11168         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
11169         (vect_transform_loop): Likewise.
11170         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11172 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11174         * tree-vectorizer.h (vect_schedule_slp): Return void.
11175         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11176         (vect_schedule_slp): Likewise.
11178 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11180         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
11181         argument.
11182         (vect_transform_loop): Update calls accordingly.  Schedule SLP
11183         instances before the main loop, if any exist.
11185 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11187         PR tree-optimization/86749
11188         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
11189         If the lhs is used in a COND_EXPR, check that it is being used
11190         as the "then" or "else" value.
11192 2018-08-01  Tom de Vries  <tdevries@suse.de>
11194         PR target/86800
11195         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
11196         speculation_safe_value_not_needed.
11198 2018-08-01  Richard Biener  <rguenther@suse.de>
11200         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
11201         as base and offset.
11203 2018-08-01  Martin Liska  <mliska@suse.cz>
11205         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
11206         format how successful transformation is dumped.
11207         (gimple_mod_pow2_value_transform): Likewise.
11208         (gimple_mod_subtract_transform): Likewise.
11209         (gimple_stringops_transform): Likewise.
11211 2018-08-01  Martin Liska  <mliska@suse.cz>
11213         PR value-prof/35543
11214         * value-prof.c (interesting_stringop_to_profile_p):
11215         Simplify the code and add BUILT_IN_MEMMOVE.
11216         (gimple_stringops_transform): Likewise.
11218 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
11220         * config/aarch64/aarch64-simd.md
11221         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
11222         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
11223         use GPI iterator instead of SI mode.
11225 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
11227         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
11228         rs6000_speculation_barrier.
11229         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
11230         new barrier pattern name.
11232 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
11234         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
11235         (speculation_barrier): New insn.
11237 2018-08-01  Richard Biener  <rguenther@suse.de>
11239         PR bootstrap/86724
11240         * graphite.h: Include isl/id.h and isl/space.h to allow build
11241         with ISL 0.20.
11243 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
11245         PR target/86651
11246         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11247         mode for COFF targets.
11248         * defaults.h (TARGET_COFF): Define.
11249         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
11250         TARGET_COFF): Define.
11251         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
11252         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
11253         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
11255 2018-07-31  Alexandre Oliva <oliva@adacore.com>
11257         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
11258         * gimple-streamer-out.c (output_bb): Save it.
11259         * lto-streamer-in.c (input_struct_function_base): Restore
11260         instance discriminator if available.  Create map on demand.
11261         * lto-streamer-out.c (output_struct_function_base): Save it if
11262         available.
11263         * final.c (decl_to_instance_map): Document LTO strategy.
11265 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
11266             Olivier Hainque  <hainque@adacore.com>
11268         * debug.h (decl_to_instance_map_t): New type.
11269         (decl_to_instance_map): Declare.
11270         (maybe_create_decl_to_instance_map): New inline function.
11271         * final.c (bb_discriminator, last_bb_discriminator): New statics,
11272         to track basic block discriminators.
11273         (final_start_function_1): Initialize them.
11274         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
11275         bb_discriminator.
11276         (decl_to_instance_map): New variable.
11277         (map_decl_to_instance, maybe_set_discriminator): New functions.
11278         (notice_source_line): Set discriminator.
11280 2018-07-31  Ian Lance Taylor  <iant@golang.org>
11282         * targhooks.c (default_have_speculation_safe_value): Add
11283         ATTRIBUTE_UNUSED.
11285 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
11287         * dump-context.h: Include "pretty-print.h".
11288         (dump_context::refresh_dumps_are_enabled): New decl.
11289         (dump_context::emit_item): New decl.
11290         (class dump_context): Add fields "m_test_pp" and
11291         "m_test_pp_flags".
11292         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
11293         (temp_dump_context::get_dumped_text): New decl.
11294         (class temp_dump_context): Add field "m_pp".
11295         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
11296         (dump_context::refresh_dumps_are_enabled): ...and add a test for
11297         m_test_pp.
11298         (set_dump_file): Update for above change.
11299         (set_alt_dump_file): Likewise.
11300         (dump_loc): New overload, taking a pretty_printer *.
11301         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
11302         to any test pretty-printer.
11303         (make_item_for_dump_gimple_stmt): New function, adapted from
11304         optinfo::add_gimple_stmt.
11305         (dump_context::dump_gimple_stmt): Call it, and use the result,
11306         eliminating the direct usage of dump_file and alt_dump_file in
11307         favor of indirectly using them via emit_item.
11308         (make_item_for_dump_gimple_expr): New function, adapted from
11309         optinfo::add_gimple_expr.
11310         (dump_context::dump_gimple_expr): Call it, and use the result,
11311         eliminating the direct usage of dump_file and alt_dump_file in
11312         favor of indirectly using them via emit_item.
11313         (make_item_for_dump_generic_expr): New function, adapted from
11314         optinfo::add_tree.
11315         (dump_context::dump_generic_expr): Call it, and use the result,
11316         eliminating the direct usage of dump_file and alt_dump_file in
11317         favor of indirectly using them via emit_item.
11318         (make_item_for_dump_printf_va): New function, adapted from
11319         optinfo::add_printf_va.
11320         (make_item_for_dump_printf): New function.
11321         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
11322         and use the result, eliminating the direct usage of dump_file and
11323         alt_dump_file in favor of indirectly using them via emit_item.
11324         (make_item_for_dump_dec): New function.
11325         (dump_context::dump_dec): Call it, and use the result,
11326         eliminating the direct usage of dump_file and alt_dump_file in
11327         favor of indirectly using them via emit_item.
11328         (make_item_for_dump_symtab_node): New function, adapted from
11329         optinfo::add_symtab_node.
11330         (dump_context::dump_symtab_node): Call it, and use the result,
11331         eliminating the direct usage of dump_file and alt_dump_file in
11332         favor of indirectly using them via emit_item.
11333         (dump_context::begin_scope): Reimplement, avoiding direct usage
11334         of dump_file and alt_dump_file in favor of indirectly using them
11335         via emit_item.
11336         (dump_context::emit_item): New member function.
11337         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
11338         Set up test pretty-printer on the underlying context.  Call
11339         refresh_dumps_are_enabled.
11340         (temp_dump_context::~temp_dump_context): Call
11341         refresh_dumps_are_enabled.
11342         (temp_dump_context::get_dumped_text): New member function.
11343         (selftest::verify_dumped_text): New function.
11344         (ASSERT_DUMPED_TEXT_EQ): New macro.
11345         (selftest::test_capture_of_dump_calls): Run all tests twice, with
11346         and then without optinfo enabled.  Add uses of
11347         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
11348         * dumpfile.h: Update comment for the dump_* API.
11349         * optinfo-emit-json.cc
11350         (selftest::test_building_json_from_dump_calls): Update for new
11351         param for temp_dump_context ctor.
11352         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
11353         and "m_owned" field.
11354         (optinfo_item::~optinfo_item): Likewise.
11355         (optinfo::add_item): New member function.
11356         (optinfo::emit): Update comment.
11357         (optinfo::add_string): Delete.
11358         (optinfo::add_printf): Delete.
11359         (optinfo::add_printf_va): Delete.
11360         (optinfo::add_gimple_stmt): Delete.
11361         (optinfo::add_gimple_expr): Delete.
11362         (optinfo::add_tree): Delete.
11363         (optinfo::add_symtab_node): Delete.
11364         (optinfo::add_dec): Delete.
11365         * optinfo.h (class dump_context): New forward decl.
11366         (optinfo::add_item): New decl.
11367         (optinfo::add_string): Delete.
11368         (optinfo::add_printf): Delete.
11369         (optinfo::add_printf_va): Delete.
11370         (optinfo::add_gimple_stmt): Delete.
11371         (optinfo::add_gimple_expr): Delete.
11372         (optinfo::add_tree): Delete.
11373         (optinfo::add_symtab_node): Delete.
11374         (optinfo::add_dec): Delete.
11375         (optinfo::add_poly_int): Delete.
11376         (optinfo_item::optinfo_item): Remove "owned" param.
11377         (class optinfo_item): Remove field "m_owned".
11379 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11381         PR middle-end/86705
11382         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
11383         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
11384         requested variable alignment.
11385         (expand_one_ssa_partition): Likewise.
11386         (expand_one_var): Likewise.
11388 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11390         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
11391         to speculation_safe_value_not_needed.
11393 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11395         * targhooks.h (speculation_safe_value_not_needed): New prototype.
11396         * targhooks.c (speculation_safe_value_not_needed): New function.
11397         * target.def (have_speculation_safe_value): Update documentation.
11398         * doc/tm.texi: Regenerated.
11400 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11402         * config/aarch64/iterators.md (ALLI_TI): New iterator.
11403         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
11404         expand.
11405         (despeculate_copy<ALLI:mode>_insn): New insn.
11406         (despeculate_copyti_insn): New insn.
11407         (despeculate_simple<ALLI:mode>): New insn
11408         (despeculate_simpleti): New insn.
11409         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
11410         function.
11411         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
11412         aarch64_speculation_safe_value.
11413         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
11415 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11417         * config/aarch64/aarch64-speculation.cc: New file.
11418         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
11419         before pass_reorder_blocks.
11420         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
11421         prototype.
11422         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
11423         X14 and X15 when tracking speculation.
11424         * config/aarch64/aarch64.md (register name constants): Add
11425         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
11426         (unspec): Add UNSPEC_SPECULATION_TRACKER.
11427         (speculation_barrier): New insn attribute.
11428         (cmp<mode>): Allow SP in comparisons.
11429         (speculation_tracker): New insn.
11430         (speculation_barrier): Add speculation_barrier attribute.
11431         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
11432         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
11433         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
11435 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11437         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
11438         aarch64_track_speculation is true.
11439         (tb<optab><mode>1): Likewise.
11440         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
11441         generate CB[N]Z when tracking speculation.
11442         (aarch64_split_compare_and_swap): Likewise.
11443         (aarch64_split_atomic_op): Likewise.
11445 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11447         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
11449 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11451         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
11452         (speculation_barrier): New insn.
11454 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11456         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
11457         * config/arm/arm.md (speculation_barrier): New expand.
11458         (speculation_barrier_insn): New pattern.
11460 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11462         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
11463         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
11464         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
11465         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
11466         list.
11467         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
11468         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
11469         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
11470         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
11471         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
11472         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
11473         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
11474         * builtins.c (expand_speculation_safe_value): New function.
11475         (expand_builtin): Call it.
11476         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
11477         * doc/extend.texi: Document __builtin_speculation_safe_value.
11478         * doc/md.texi: Document "speculation_barrier" pattern.
11479         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
11480         TARGET_HAVE_SPECULATION_SAFE_VALUE.
11481         * doc/tm.texi: Regenerated.
11482         * target.def (have_speculation_safe_value, speculation_safe_value): New
11483         hooks.
11484         * targhooks.c (default_have_speculation_safe_value): New function.
11485         (default_speculation_safe_value): New function.
11486         * targhooks.h (default_have_speculation_safe_value): Add prototype.
11487         (default_speculation_safe_value): Add prototype.
11489 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
11491         * dump-context.h (dump_context::dump_loc): New decl.
11492         * dumpfile.c (dump_context::dump_loc): New member function.
11493         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
11494         and dump_gimple_stmt.
11495         (dump_context::dump_gimple_expr_loc): Likewise, using
11496         dump_gimple_expr.
11497         (dump_context::dump_generic_expr_loc): Likewise, using
11498         dump_generic_expr.
11499         (dump_context::dump_printf_loc_va): Likewise, using
11500         dump_printf_va.
11501         (dump_context::begin_scope): Explicitly using the global function
11502         "dump_loc", rather than the member function.
11504 2018-07-31  Martin Sebor  <msebor@redhat.com>
11506         PR tree-optimization/86741
11507         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
11509 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
11511         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
11512         depend on whether prefetch instructions will be emitted or not.
11513         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
11514         will be emitted or not.
11515         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
11516         (TARGET_SETMEM_PFD): New macros.
11518 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11520         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
11521         (NULL_STMT_VEC_INFO): Delete.
11522         (stmt_vec_info::operator*): Likewise.
11523         (stmt_vec_info::operator gimple *): Likewise.
11524         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
11525         of NULL_STMT_VEC_INFO.
11526         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11527         (vect_reassociating_reduction_p): Likewise.
11528         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11529         (vectorizable_store): Likewise.
11530         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
11531         (vec_info::free_stmt_vec_infos): Likewise.
11533 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11535         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
11536         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
11538 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11540         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
11541         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
11542         (vec_info::free_stmt_vec_info): New private member functions.
11543         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
11544         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
11545         * tree-parloops.c (gather_scalar_reductions): Remove calls to
11546         set_stmt_vec_info_vec and free_stmt_vec_infos.
11547         * tree-vect-loop.c (_loop_vec_info): Remove call to
11548         set_stmt_vec_info_vec.
11549         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
11550         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
11551         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
11552         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
11553         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
11554         assignments in {vec_info::,}new_stmt_vec_info that are redundant
11555         with the clearing in the xcalloc.
11556         (stmt_vec_info_vec): Delete.
11557         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
11558         (vectorize_loops): Likewise.
11559         (vec_info::~vec_info): Remove argument from call to
11560         free_stmt_vec_infos.
11561         (vec_info::add_stmt): Remove vinfo argument from call to
11562         new_stmt_vec_info.
11564 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11566         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
11567         rather than a gimple stmt.
11568         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
11569         information for pattern statements when passed the original
11570         statement; instead wait to be passed the pattern statement itself.
11571         Don't call set_vinfo_for_stmt here.
11572         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
11573         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
11574         stmt_vec_infos here.
11575         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
11576         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
11577         stmt_vec_infos entry.
11579 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11581         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
11582         * tree-vectorizer.c (vec_info::replace_stmt): New function.
11583         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
11584         * tree-vect-stmts.c (vectorizable_call): Likewise.
11585         (vectorizable_simd_clone_call): Likewise.
11587 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11589         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
11590         * tree-vectorizer.c (vec_info::remove_stmt): New function.
11591         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
11592         * tree-vect-loop.c (vect_transform_loop): Likewise.
11593         * tree-vect-slp.c (vect_schedule_slp): Likewise.
11594         * tree-vect-stmts.c (vect_remove_stores): Likewise.
11596 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11598         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
11599         (vect_dr_stmt): Delete.
11600         * tree-vectorizer.c (vec_info::lookup_dr): New function.
11601         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
11602         of DR_VECT_AUX.
11603         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11604         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
11605         (vect_verify_datarefs_alignment, vect_peeling_supportable)
11606         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11607         (vect_analyze_data_refs): Likewise.
11608         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
11609         argument.
11610         (vect_find_same_alignment_drs): Likewise.
11611         (vect_slp_analyze_node_dependences): Update calls accordingly.
11612         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
11613         instead of DR_VECT_AUX.
11614         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
11615         of a vector data references.  Use vec_info::lookup_dr instead of
11616         DR_VECT_AUX.
11617         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
11618         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
11619         instead of DR_VECT_AUX.
11621 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11623         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
11624         dr_vec_info.
11625         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11626         accordingly.
11627         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11628         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11629         (vect_gen_prolog_loop_niters): Likewise.
11631 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11633         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
11634         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
11635         (vect_known_alignment_in_bytes, vect_dr_behavior)
11636         (vect_get_scalar_dr_size): Take references as dr_vec_infos
11637         instead of data_references.  Update calls to other routines for
11638         which the same change has been made.
11639         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
11640         dr_vec_infos instead of stmt_vec_infos.
11641         (vect_analyze_data_ref_dependence): Update call accordingly.
11642         (vect_slp_analyze_data_ref_dependence)
11643         (vect_record_base_alignments): Use DR_VECT_AUX.
11644         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
11645         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
11646         (vector_alignment_reachable_p, vect_get_data_access_cost)
11647         (vect_peeling_supportable, vect_analyze_group_access_1)
11648         (vect_analyze_group_access, vect_analyze_data_ref_access)
11649         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
11650         (vect_compile_time_alias, vect_small_gap_p)
11651         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
11652         (vect_supportable_dr_alignment): Take references as dr_vec_infos
11653         instead of data_references.  Update calls to other routines for
11654         which the same change has been made.
11655         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
11656         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11657         (vect_slp_analyze_and_verify_node_alignment)
11658         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11659         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11660         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
11661         above changes.
11662         (_vect_peel_info::dr): Replace with...
11663         (_vect_peel_info::dr_info): ...this new field.
11664         (vect_peeling_hash_get_most_frequent)
11665         (vect_peeling_hash_choose_best_peeling): Update accordingly.
11666         (vect_peeling_hash_get_lowest_cost):
11667         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
11668         routines for which the same change has been made.
11669         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
11670         data_reference.
11671         * tree-vect-loop-manip.c (get_misalign_in_elems)
11672         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
11673         above changes.
11674         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11675         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
11676         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
11677         (get_group_load_store_type, get_negative_load_store_type)
11678         (vect_get_data_ptr_increment, vectorizable_store)
11679         (vectorizable_load): Likewise.
11680         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
11681         Update calls to other routines for which the same change has been made.
11683 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11685         * tree-vectorizer.h (vec_info::move_dr): New member function.
11686         (dataref_aux): Rename to...
11687         (dr_vec_info): ...this and add "dr" and "stmt" fields.
11688         (_stmt_vec_info::dr_aux): Update accordingly.
11689         (_stmt_vec_info::data_ref_info): Delete.
11690         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11691         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11692         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11693         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11694         of data_ref.
11695         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
11696         (STMT_VINFO_DR_INFO): New macro.
11697         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11698         (set_dr_misalignment): Update after rename of dataref_aux.
11699         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
11700         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11701         initialization of STMT_VINFO_DATA_REF.
11702         * tree-vectorizer.c (vec_info::move_dr): New function.
11703         * tree-vect-patterns.c (vect_recog_bool_pattern)
11704         (vect_recog_mask_conversion_pattern)
11705         (vect_recog_gather_scatter_pattern): Use it.
11706         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11707         the "dr" and "stmt" fields of dr_vec_info instead of
11708         STMT_VINFO_DATA_REF.
11710 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11712         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11713         (is_pattern_stmt_p): Use it.
11714         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11715         on pattern statements.
11717 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11719         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11720         original stmt as a stmt_vec_info rather than a gimple stmt.
11721         (vect_pattern_recog_1): Take the statement directly as a
11722         stmt_vec_info, rather than via a gimple_stmt_iterator.
11723         Update call to vect_mark_pattern_stmts.
11724         (vect_pattern_recog): Update calls accordingly.
11726 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11728         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11729         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11730         a vect_def_type for the first argument.
11731         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11732         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
11733         operand if it isn't defined by a vectorized statement.
11734         (vect_build_gather_load_calls): Remove the mask_dt argument and
11735         update calls to vect_get_vec_def_for_stmt_copy.
11736         (vectorizable_bswap): Likewise the dt argument.
11737         (vectorizable_call): Update calls to vectorizable_bswap and
11738         vect_get_vec_def_for_stmt_copy.
11739         (vectorizable_simd_clone_call, vectorizable_assignment)
11740         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11741         (vectorizable_comparison): Update calls to
11742         vect_get_vec_def_for_stmt_copy.
11743         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
11744         vect_is_simple_use.
11745         (vect_get_loop_based_defs): Remove dt argument and update call
11746         to vect_get_vec_def_for_stmt_copy.
11747         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11748         and vect_get_vec_def_for_stmt_copy.
11749         (vectorizable_load): Update calls to vect_build_gather_load_calls
11750         and vect_get_vec_def_for_stmt_copy.
11751         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11752         (vectorizable_reduction, vectorizable_live_operation): Update calls
11753         to vect_get_vec_def_for_stmt_copy.
11755 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11757         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11758         and gimple stmt arguments with a stmt_vec_info.
11759         (vect_record_base_alignments): Update calls accordingly.
11760         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11761         and gimple stmt arguments with a stmt_vec_info.
11762         (vect_build_slp_tree_1): Remove vinfo argument and update call
11763         to vect_record_max_nunits.
11764         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11765         and vect_record_max_nunits.
11767 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11769         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11770         file and take a stmt_vec_info instead of a gimple stmt.
11771         (supportable_widening_operation, vect_finish_replace_stmt)
11772         (vect_finish_stmt_generation, vect_get_store_rhs)
11773         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11774         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11775         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11776         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11777         (vect_create_data_ref_ptr, bump_vector_ptr)
11778         (vect_permute_store_chain, vect_setup_realignment)
11779         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11780         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11781         (vectorizable_reduction, vectorizable_induction)
11782         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11783         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11784         than gimple stmts as arguments.
11785         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11786         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11787         (can_group_stmts_p, vect_check_gather_scatter)
11788         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11789         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11790         (vect_permute_load_chain, vect_shift_permute_load_chain)
11791         (vect_transform_grouped_load)
11792         (vect_record_grouped_load_vectors): Likewise.
11793         * tree-vect-loop.c (vect_fixup_reduc_chain)
11794         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11795         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11796         (vectorizable_reduction, vectorizable_induction)
11797         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11798         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11799         (vect_get_load_store_mask): Likewise.
11800         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11801         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11802         * tree-vect-stmts.c (vect_mark_relevant)
11803         (is_simple_and_all_uses_invariant)
11804         (exist_non_indexing_operands_for_use_p, process_use)
11805         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11806         (vect_get_vec_def_for_operand, vect_get_vec_defs)
11807         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11808         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11809         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11810         (get_negative_load_store_type, get_load_store_type)
11811         (vect_check_load_store_mask, vect_check_store_rhs)
11812         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11813         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11814         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11815         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11816         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11817         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11818         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11819         (supportable_widening_operation): Likewise.
11821 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11823         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11824         a stmt_vec_info instead of a gcall.
11825         (vect_check_gather_scatter): Update call accordingly.
11826         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11827         of a gphi.
11828         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11829         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11830         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11831         instead of a gimple stmt.
11832         (vect_transform_loop): Update calls accordingly.
11833         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11834         stmt_vec_infos instead of gimple stmts.
11835         (vect_analyze_slp_instance): Update use accordingly.
11836         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11837         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11838         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11839         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11840         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11841         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11842         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11843         down to subroutines.
11845 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11847         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11848         of the worklist from a vector of gimple stmts to a vector of
11849         stmt_vec_infos.
11850         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11851         (vect_mark_stmts_to_be_vectorized): Likewise
11853 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11855         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11856         statement before passing it to vect_analyze_stmt.
11857         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11858         the chain of phi vector definitions.  Track the exit phi via its
11859         stmt_vec_info.
11860         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11861         STMT_VINFO_REDUC_DEF.
11862         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11863         stmt_vec_infos to handle the statement chains.
11864         (vect_get_slp_defs): Record the first statement in the node
11865         using a stmt_vec_info.
11866         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11867         statements here and pass their stmt_vec_info down to subroutines.
11868         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11869         down to vect_finish_stmt_generation.
11870         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11871         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11872         stmt_vec_infos to subroutines.
11873         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11874         chains.
11876 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11878         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11879         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11880         (vect_permute_store_chain, vect_setup_realignment)
11881         (vect_permute_load_chain, vect_shift_permute_load_chain)
11882         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11883         stmts internally, and when passing values to other vectorizer routines.
11884         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11885         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11886         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11887         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11888         (vectorizable_reduction, vectorizable_induction)
11889         (vectorizable_live_operation, vect_transform_loop_stmt)
11890         (vect_transform_loop): Likewise.
11891         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11892         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11893         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11894         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11895         (vect_slp_analyze_node_operations_1): Likewise.
11896         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11897         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11898         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11899         (vect_finish_stmt_generation_1, get_group_load_store_type)
11900         (get_load_store_type, vect_build_gather_load_calls)
11901         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11902         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11903         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11904         (vectorizable_store, vectorizable_load, vectorizable_condition)
11905         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11906         (supportable_widening_operation): Likewise.
11907         (vect_get_vector_types_for_stmt): Likewise.
11908         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11910 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11912         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11913         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11914         (vect_permute_store_chain, vect_permute_load_chain)
11915         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11916         repeated stmt_vec_info lookups.
11917         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11918         (vect_update_ivs_after_vectorizer): Likewise.
11919         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11920         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11921         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11922         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11923         (vect_bb_slp_scalar_cost): Likewise.
11924         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11926 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11928         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11929         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11930         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11931         to get gassigns and gcalls, rather than operating on generc gimple
11932         stmts.
11933         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11934         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11935         (vectorizable_load, vect_analyze_stmt): Likewise.
11936         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11938 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11940         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11941         return stmt_vec_infos rather than gimple stmts.  Do not accept
11942         null arguments.
11943         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11944         of a gimple stmt.
11945         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11946         Update use of get_later_stmt.
11947         (vect_get_constant_vectors): Update call accordingly.
11948         (vect_schedule_slp_instance): Likewise
11949         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11950         (vect_slp_analyze_instance_dependence): Likewise.
11951         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11953 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11955         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11956         (stmt_info_for_cost::stmt_info): ...this new field.
11957         (add_stmt_costs): Update accordingly.
11958         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11959         (vect_get_known_peeling_cost): Likewise.
11960         (vect_estimate_min_profitable_iters): Likewise.
11961         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11963 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11965         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11966         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11967         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11968         accordingly.
11969         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11971 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11973         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11974         a gimple stmt to a stmt_vec_info.
11975         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11977 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11979         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11980         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11981         (_loop_vec_info::reduction_chains): Likewise.
11982         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11983         accordingly.
11984         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11988         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11989         a gimple stmt to a stmt_vec_info.
11990         (_stmt_vec_info::next_element): Likewise.
11991         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11992         (vect_slp_analyze_and_verify_node_alignment)
11993         (vect_analyze_group_access_1, vect_analyze_group_access)
11994         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11995         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11996         (vect_supportable_dr_alignment): Update accordingly.
11997         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11998         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11999         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
12000         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12001         * tree-vect-slp.c (vect_build_slp_tree_1)
12002         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
12003         (vect_split_slp_store_group, vect_analyze_slp_instance)
12004         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
12005         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
12006         (get_group_load_store_type, get_load_store_type)
12007         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
12008         (vect_transform_stmt, vect_remove_stores): Likewise.
12010 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12012         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
12013         than a gimple stmt.
12014         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12015         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
12016         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
12017         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
12018         (vector_alignment_reachable_p, vect_get_data_access_cost)
12019         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
12020         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
12021         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
12022         (vect_analyze_group_access_1, vect_analyze_group_access)
12023         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
12024         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
12025         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
12026         result of vect_dr_stmt and use the stmt_vec_info instead of
12027         the associated gimple stmt.
12028         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12029         (vect_gen_prolog_loop_niters): Likewise.
12030         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12032 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12034         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
12035         to a vec<stmt_vec_info>.
12036         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
12037         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
12038         vec<stmt_vec_info>.
12039         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
12040         to a vec<stmt_vec_info>.
12041         (bst_traits::value_type, bst_traits::value_type): Likewise.
12042         (bst_traits::hash): Update accordingly.
12043         (vect_get_and_check_slp_defs): Change the stmts parameter from
12044         a vec<gimple *> to a vec<stmt_vec_info>.
12045         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
12046         (vect_build_slp_tree): Likewise.
12047         (vect_build_slp_tree_2): Likewise.  Update uses of
12048         SLP_TREE_SCALAR_STMTS.
12049         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
12050         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
12051         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
12052         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
12053         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
12054         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
12055         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
12056         (vect_get_constant_vectors, vect_get_slp_defs)
12057         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
12058         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
12059         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
12060         instead of gimple stmts.
12061         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
12062         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
12063         (vect_slp_analyze_instance_dependence): Update uses of
12064         SLP_TREE_SCALAR_STMTS.
12065         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12066         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
12067         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
12068         (get_initial_defs_for_reduction): Likewise.
12069         (vect_create_epilog_for_reduction): Likewise.
12070         (vectorize_fold_left_reduction): Likewise.
12071         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
12072         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
12073         (can_vectorize_live_stmts): Likewise.
12075 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12077         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
12078         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12079         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
12080         than gimple stmts.
12081         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
12082         of a gimple stmt.
12083         (gather_scalar_reductions): Update after above interface changes.
12084         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
12085         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
12086         than gimple stmts.
12087         (vect_force_simple_reduction): Likewise.
12088         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
12089         LOOP_VINFO_REDUCTIONS.
12090         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12092 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12094         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
12095         a gimple stmt to a stmt_vec_info.
12096         * tree-vect-loop.c (vect_active_double_reduction_p)
12097         (vect_force_simple_reduction, vectorizable_reduction): Update
12098         accordingly.
12100 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12102         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
12103         vec<gimple *> to a vec<stmt_vec_info>.
12104         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
12105         the reduction_phis argument from a vec<gimple *> to a
12106         vec<stmt_vec_info>.
12107         (vectorizable_reduction): Likewise the phis local variable that
12108         is passed to vect_create_epilog_for_reduction.  Update for new type
12109         of SLP_TREE_VEC_STMTS.
12110         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
12111         (vectorizable_live_operation): Likewise.
12112         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
12113         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
12115 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12117         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
12118         a gimple stmt to a stmt_vec_info.
12119         (vectorizable_condition, vectorizable_live_operation)
12120         (vectorizable_reduction, vectorizable_induction): Pass back the
12121         vectorized statement as a stmt_vec_info.
12122         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
12123         use of STMT_VINFO_VEC_STMT.
12124         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
12125         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
12126         as stmt_vec_infos rather than gimple stmts.
12127         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
12128         to a stmt_vec_info.
12129         (vectorizable_live_operation): Likewise.
12130         (vectorizable_reduction, vectorizable_induction): Likewise,
12131         updating use of STMT_VINFO_VEC_STMT.
12132         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
12133         of STMT_VINFO_VEC_STMT.
12134         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
12135         (vectorizable_simd_clone_call, vectorizable_conversion)
12136         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12137         (vectorizable_store, vectorizable_load, vectorizable_condition)
12138         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
12139         from a gimple stmt to a stmt_vec_info.
12140         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
12141         pointer to a stmt_vec_info to the vectorizable_* routines.
12143 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12145         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
12146         a gimple stmt to a stmt_vec_info.
12147         (is_pattern_stmt_p): Update accordingly.
12148         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
12149         (vect_record_grouped_load_vectors): Likewise.
12150         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
12151         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
12152         (vect_model_reduction_cost): Likewise.
12153         (vect_create_epilog_for_reduction): Likewise.
12154         (vectorizable_reduction, vectorizable_induction): Likewise.
12155         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12156         Return the stmt_vec_info for the pattern statement.
12157         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
12158         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
12159         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
12160         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
12161         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
12162         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
12163         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
12164         (free_stmt_vec_info, vect_is_simple_use): Likewise.
12166 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12168         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
12169         (vect_finish_stmt_generation): Likewise.
12170         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
12171         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
12172         (vect_build_gather_load_calls): Use the return value of the above
12173         functions instead of a separate call to vinfo_for_stmt.  Use narrow
12174         scopes for the input gimple stmt and wider scopes for the associated
12175         stmt_vec_info.  Use vec_info::lookup_def when setting these
12176         stmt_vec_infos from an SSA_NAME definition.
12177         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12178         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12179         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12180         (vectorizable_store, vectorizable_load, vectorizable_condition)
12181         (vectorizable_comparison): Likewise.
12182         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
12183         (vectorizable_reduction): Likewise.
12185 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12187         * tree-vectorizer.h (vect_is_simple_use): Add an optional
12188         stmt_vec_info * parameter before the optional gimple **.
12189         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12190         (process_use, vect_get_vec_def_for_operand_1): Update callers.
12191         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
12192         * tree-vect-loop.c (vectorizable_reduction): Likewise.
12193         (vectorizable_live_operation): Likewise.
12194         * tree-vect-patterns.c (type_conversion_p): Likewise.
12195         (vect_look_through_possible_promotion): Likewise.
12196         (vect_recog_rotate_pattern): Likewise.
12197         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12199 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12201         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
12202         a typedef to a wrapper class.
12203         (NULL_STMT_VEC_INFO): New macro.
12204         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
12205         (stmt_vec_info::operator*): New function.
12206         (stmt_vec_info::operator gimple *): Likewise.
12207         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
12208         (add_stmt_costs): Likewise.
12209         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
12210         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
12211         (vect_get_known_peeling_cost): Likewise.
12212         (vect_estimate_min_profitable_iters): Likewise.
12213         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12214         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
12215         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
12216         (vectorizable_store, free_stmt_vec_infos): Likewise.
12217         (new_stmt_vec_info): Change return type of xcalloc to
12218         _stmt_vec_info *.
12220 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12222         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
12223         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
12224         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
12225         a single_imm_use-based sequence.
12226         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
12228 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12230         * tree-vectorizer.h (vec_info::lookup_def): Declare.
12231         * tree-vectorizer.c (vec_info::lookup_def): New function.
12232         * tree-vect-patterns.c (vect_get_internal_def): Use it.
12233         (vect_widened_op_tree): Likewise.
12234         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12235         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
12236         (vectorizable_reduction): Likewise.
12237         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
12238         of a gimple *.
12239         (vect_is_slp_reduction): Update calls accordingly.  Use
12240         vec_info::lookup_def.
12241         (vect_is_simple_reduction): Likewise
12242         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
12244 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12246         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
12247         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
12248         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
12249         of vinfo_for_stmt.
12250         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
12251         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
12252         (vect_update_vf_for_slp, vect_analyze_loop_operations)
12253         (vect_is_slp_reduction, vectorizable_induction)
12254         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
12255         * tree-vect-patterns.c (vect_init_pattern_stmt):
12256         (vect_determine_min_output_precision_1, vect_determine_precisions)
12257         (vect_pattern_recog): Likewise.
12258         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
12259         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
12260         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
12261         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
12262         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
12263         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
12264         info field from a loop to a loop_vec_info.
12266 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12268         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
12269         (vec_info::add_stmt): Declare.
12270         * tree-vectorizer.c (vec_info::add_stmt): New function.
12271         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
12272         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
12273         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
12274         (vectorizable_induction): Likewise.
12275         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
12276         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
12277         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
12278         (vectorizable_load): Likewise.
12279         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12280         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
12281         (vect_recog_gather_scatter_pattern): Likewise.
12282         (append_pattern_def_seq): Likewise.  Remove a check that is
12283         performed by add_stmt itself.
12285 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12287         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
12288         which make_ssa_name was called with new_stmt before new_stmt
12289         had been created.
12291 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12293         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
12294         split out from...
12295         (vect_is_slp_reduction): ...here...
12296         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
12297         that are already known to be false.
12299 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12301         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
12302         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
12303         STMT_VINFO_NUM_SLP_USES when it's true.
12304         (vect_free_slp_instance): Add a final_p parameter and pass it to
12305         vect_free_slp_tree.
12306         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
12307         (vect_analyze_slp_instance): Likewise.
12308         (vect_slp_analyze_operations): Likewise.
12309         (vect_slp_analyze_bb_1): Likewise.
12310         * tree-vectorizer.c (vec_info): Likewise.
12311         * tree-vect-loop.c (vect_transform_loop): Likewise.
12313 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12315         * tree-vect-loop.c (vectorizable_reduction): Assert that the
12316         function is not called for second and subsequent members of
12317         a reduction group.
12319 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12321         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
12322         cases for nested loops from here to ...
12323         (vect_create_epilog_for_reduction): ...here.  Only call
12324         vect_is_simple_use for inner-loop reductions.
12326 2018-07-31  Martin Liska  <mliska@suse.cz>
12328         PR gcov-profile/85338
12329         PR gcov-profile/85350
12330         PR gcov-profile/85372
12331         * profile.c (struct location_triplet): New.
12332         (struct location_triplet_hash): Likewise.
12333         (output_location): Do not output a BB that
12334         is already recorded for a line.
12335         (branch_prob): Use streamed_locations.
12337 2018-07-31  Martin Liska  <mliska@suse.cz>
12339         PR gcov-profile/85370
12340         * coverage.c (coverage_begin_function): Do not mark target
12341         clones as artificial functions.
12343 2018-07-31  Martin Liska  <mliska@suse.cz>
12345         PR gcov-profile/83813
12346         PR gcov-profile/84758
12347         PR gcov-profile/85217
12348         PR gcov-profile/85332
12349         * profile.c (branch_prob): Do not record GOTO expressions
12350         for GIMPLE statements which locations are already streamed.
12352 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12354         * gcc.c (handle_spec_function): Accept a soft_matched_part
12355         argument, as do_spec_1.  Pass it down to ...
12356         (eval_spec_function): Accept a soft_matched_part argument,
12357         and pass it down to ...
12358         (do_spec_2): Accept a soft_matched_part argument, and pass
12359         it down to do_spec_1.
12360         (do_spec_1): Pass soft_matched_part to handle_spec_function.
12361         (handle_braces): Update call to handle_spec_function.
12362         (driver::set_up_specs): Update calls to do_spec_2.
12363         (compare_debug_dump_opt_spec_function): Likewise.
12364         (compare_debug_self_opt_spec_function): Likewise.
12366 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12368         * common.opt (nolibc): New option.
12369         * doc/invoke.texi (Link Options): Document it.
12370         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
12371         * config/alpha/linux.h: Likewise.
12372         * config/arc/elf.h: Likewise.
12373         * config/arm/uclinux-elf.h: Likewise.
12374         * config/arm/unknown-elf.h: Likewise.
12375         * config/avr/avrlibc.h: Likewise.
12376         * config/bfin/bfin.h: Likewise.
12377         * config/bfin/linux.h: Likewise.
12378         * config/bfin/uclinux.h: Likewise.
12379         * config/darwin.h: Likewise.
12380         * config/darwin10.h: Likewise.
12381         * config/darwin12.h: Likewise.
12382         * config/gnu-user.h: Likewise.
12383         * config/lm32/uclinux-elf.h: Likewise.
12384         * config/pa/pa-hpux11.h: Likewise.
12385         * config/pa/pa64-hpux.h: Likewise.
12386         * config/sparc/sparc.h: Likewise.
12388 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12390         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
12391         undefined variables.
12393 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
12395         PR target/86640
12396         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
12397         instead of GEN_INT.
12399 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12401         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
12402         terminated string literal.
12404 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
12406         PR rtl-optimization/85160
12407         * combine.c (is_just_move): New function.
12408         (try_combine): Allow combining two instructions into two if neither of
12409         the original instructions was a move.
12411 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
12413         PR target/86673
12414         * doc/extend.texi (Global Register Variables): Discourage use of type
12415         qualifiers.
12416         (Local Register Variables): Likewise.
12418 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
12420         PR tree-optimization/86506
12421         * hwint.h (ceil_log2): Resync with hwint.c implementation.
12423 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
12425         PR target/86547
12426         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
12427         hard_regno, make sure no insn between `from` and `to` clobbers it.
12429 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
12430             Tom de Vries  <tdevries@suse.de>
12432         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
12433         (PTX_DEFAULT_RUNTIME_DIM): ... this.
12434         (nvptx_goacc_validate_dims): Set default worker and gang dims to
12435         PTX_DEFAULT_RUNTIME_DIM.
12436         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
12438 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
12440         * config/pa/pa.c (pa_output_addr_vec): Align address table.
12441         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
12442         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
12444 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
12446         * config/rs6000/constraints.md (wG constraint): Delete, no longer
12447         used.
12448         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
12449         predicate to reflect toc fusion has been deleted.
12450         (toc_fusion_mem_raw): Delete, no longer used.
12451         (toc_fusion_mem_wrapped): Likewise.
12452         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
12453         fusion mask bit.
12454         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
12455         Delete, no longer used.
12456         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
12457         meant to be used for toc fusion.
12458         (rs6000_debug_print_mode): Delete toc fusion debugging.
12459         (rs6000_debug_reg_global): Likewise.
12460         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
12461         fusion and secondary reload support that were never used.
12462         (rs6000_option_override_internal): Delete TOC fusion, that was only
12463         partially defined, and it did not work unless you also used the
12464         -mcmodel= switch.
12465         (rs6000_legitimate_address_p): Delete TOC fusion support.
12466         (rs6000_opt_masks): Likewise.
12467         (fusion_wrap_memory_address): Delete function, no longer used.
12468         (fusion_split_address); Delete TOC fusion support.
12469         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
12470         longer used with toc fusion being deleted.
12471         (TARGET_TOC_FUSION_FP): Likewise.
12472         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
12473         UNSPEC.
12474         (toc fusion spliter): Delete TOC fusion support.
12475         (toc_fusionload_<mode>): Likewise.
12476         (toc_fusionload_di): Likewise.
12477         (fusion_gpr_load_<mode>): Delete generator function, this insn no
12478         longer needs to be named.  Rename predicate to delete TOC fusion.
12479         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
12480         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
12481         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
12482         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
12483         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
12485 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12487         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
12488         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
12489         __int128 in built-in function prototypes.
12490         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
12491         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
12493 2018-07-27  Martin Sebor  <msebor@redhat.com>
12495         PR tree-optimization/86696
12496         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
12497         types, including enums.
12498         (handle_char_store): Be prepared for the above function to fail.
12500 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
12502         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
12503         when optimization level is lower than 2 or optimize for size.
12505 2018-07-26  Martin Sebor  <msebor@redhat.com>
12507         PR tree-optimization/86043
12508         PR tree-optimization/86042
12509         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
12510         (get_string_cst_length): Rename...
12511         (get_min_string_length): ...to this.  Add argument.
12512         (handle_char_store): Extend to handle multi-character stores by
12513         MEM_REF.
12514         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
12515         * tree.h (initializer_zerop): Add argument.
12517 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
12519         PR middle-end/86660
12520         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
12521         declare target to variables if they have always,{to,from,tofrom} map
12522         kinds.
12524 2018-07-26  Martin Liska  <mliska@suse.cz>
12526         PR lto/86548
12527         * lto-wrapper.c: Add linker_output as prefix
12528         for ltrans_output_file.
12530 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
12532         PR rtl-optimization/85805
12533         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
12534         value for hard registers if that was written in the same mode.
12536 2018-07-26  Martin Liska  <mliska@suse.cz>
12538         PR gcov-profile/86536
12539         * gcov.c (format_gcov): Use printf format %.*f directly
12540         and do not handle special values.
12542 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
12544         * common/config/arc/arc-common.c (arc_option_optimization_table):
12545         Update default optimizations for size.
12547 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
12549         * config/arc/arc.md (movsf_insn): Add short instruction selection.
12550         * config/arc/constraints.md (CfZ): New constraint.
12551         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
12552         (subsf3_fpu): Likewise.
12553         (cmpsf_fpu): Likewise.
12554         (cmpsf_fpu_uneq): Likewise.
12556 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
12558         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
12559         accesses as well.
12560         (arc_is_uncached_mem_p): uncached applies to both the variable and
12561         the pointer.
12563 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
12565         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
12566         register names.
12568 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
12570         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
12571         field "m_scopes" from vec to auto_vec.
12573 2018-07-25  Martin Liska  <mliska@suse.cz>
12575         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
12576         return type.
12578 2018-07-25  Richard Biener  <rguenther@suse.de>
12580         PR debug/86654
12581         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
12582         special wrt context_die late.
12583         (gen_subprogram_die): Re-use DIEs in local scope.
12585 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
12587         PR tree-optimization/86644
12588         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
12590 2018-07-25  Martin Liska  <mliska@suse.cz>
12592         PR middle-end/86645
12593         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
12594         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
12596 2018-07-25  Martin Liska  <mliska@suse.cz>
12598         PR sanitizer/79635
12599         * params.def: Explain ASan abbreviation and provide
12600         a documentation link.
12602 2018-07-24  Martin Sebor  <msebor@redhat.com>
12604         PR tree-optimization/86622
12605         PR tree-optimization/86532
12606         * builtins.h (string_length): Declare.
12607         * builtins.c (c_strlen): Correct handling of non-constant offsets.
12608         (check_access): Be prepared for non-constant length ranges.
12609         (string_length): Make extern.
12610         * expr.c (string_constant): Only handle the minor non-constant
12611         array index.  Use string_constant to compute the length of
12612         a generic string constant.
12614 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
12616         PR tree-optimization/86618
12617         * tree-vect-stmts.c (vectorizable_call): Don't take the address
12618         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
12620 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
12622         PR tree-optimization/86636
12623         * json.cc (json::object::set): Fix comment.  Add assertions.
12624         (json::array::append): Move here from json.h.  Add comment and an
12625         assertion.
12626         (json::string::string): Likewise.
12627         * json.h (json::array::append): Move to json.cc.
12628         (json::string::string): Likewise.
12629         * optinfo-emit-json.cc
12630         (optrecord_json_writer::impl_location_to_json): Assert that we
12631         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
12632         wrapper around it.  Expand the location once, rather than three
12633         times.
12634         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
12635         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
12636         wrappers.
12637         (optrecord_json_writer::optinfo_to_json): Likewise, in four
12638         places.  Fix some overlong lines.
12640 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
12642         * config/aarch64/aarch64-simd.md
12643         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
12644         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
12645         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
12646         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
12647         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
12648         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
12649         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
12650         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12651         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
12653 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
12655         PR middle-end/86627
12656         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
12657         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
12658         and abs_d == d, do the power of two handling if profitable.
12660 2018-07-24  Richard Biener  <rguenther@suse.de>
12662         * match.pd: Add BIT_FIELD_REF canonicalizations.
12664 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12666         PR c/86617
12667         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
12669 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12671         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
12672         terminated STRING_CST object.
12674 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12676         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
12678 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12680         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
12681         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
12682         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
12683         the elements into a register.
12684         (rs6000_split_v4si_init_di_reg): Delete.
12685         (rs6000_split_v4si_init): Delete.
12686         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12687         (vsx_init_v4si): Rewrite as a define_expand.
12689 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12691         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12692         zero_extend argument from memory): New.
12694 2018-07-22  Martin Sebor  <msebor@redhat.com>
12696         PR bootstrap/86621
12697         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12698         diagnosing calls with unknown arguments unless -Walloca-larger-than
12699         is restricted to less than PTRDIFF_MAX bytes.
12701 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
12703         * doc/gcov.texi (Invoking Gcov): Editorial changes.
12705 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12707         * pretty-print.c (text_info::set_location): Remove redundant
12708         "line_table" parameter from call to rich_location::set_range.
12710 2018-07-20  Martin Sebor  <msebor@redhat.com>
12712         PR middle-end/82063
12713         * builtins.c (expand_builtin_alloca): Adjust.
12714         * calls.c (alloc_max_size): Simplify.
12715         * cgraphunit.c (cgraph_node::expand): Adjust.
12716         * common.opt (larger_than_size, warn_frame_larger_than): Remove
12717         variables.
12718         (frame_larger_than_size): Same.
12719         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12720         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12721         Initialize.
12722         * doc/invoke.texi (GCC Command Options): Document option arguments.
12723         Explain byte-size arguments and suffixes.
12724         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12725         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12726         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12727         * doc/options.texi (UInteger): Expand.
12728         (Host_Wide_Int, ByteSize): Document new properties.
12729         * final.c (final_start_function_1): Include sizes in an error message.
12730         * function.c (frame_offset_overflow): Same.
12731         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12732         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12733         Diagnose unbounded alloca calls only for limits of less than
12734         PTRDIFF_MAX.
12735         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
12736         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
12737         for alloca(0).
12738         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
12739         only for limits of less than PTRDIFF_MAX.
12740         * langhooks-def.h (lhd_handle_option): Change function argument
12741         to HOST_WIDE_INT.
12742         * langhooks.c (lhd_handle_option): Same.
12743         * langhooks.h (handle_option): Same.
12744         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12745         ByteSize flags.
12746         (var_type, var_type_struct): Same.
12747         (var_set): Handle ByteSize flag.
12748         * optc-gen.awk: Add comments to output to ease debugging.  Make
12749         use of HOST_WIDE_INT where appropriate.
12750         * opts-gen-save.awk:  Use %lx to format unsigned long.
12751         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12752         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12753         arguments.  Parse bytes-size suffixes.
12754         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12755         (enum_value_to_arg): Same.
12756         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
12757         (handle_option): Adjust.
12758         (generate_option): Change function argument to HOST_WIDE_INT.
12759         (cmdline_handle_error): Adjust.
12760         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12761         (set_option): Change function argument to HOST_WIDE_INT.
12762         (option_enabled): Handle cl_host_wide_int.
12763         (get_option_state): Handle CLVC_SIZE.
12764         (control_warning_option): Same.
12765         * opts.c (common_handle_option): Change function argument to
12766         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
12767         OPT_Wvla_larger_than_.
12768         * opts.h (enum cl_var_type): Add an enumerator.
12769         * stor-layout.c (layout_decl): Print a more meaningful warning.
12770         * toplev.c (output_stack_usage): Adjust.
12772 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
12774         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12775         call to inline_expand_builtin_string_cmp.
12776         (expand_builtin_strcmp): Likewise.
12777         (expand_builtin_strncmp): Likewise.
12778         (inline_string_cmp): Delete the last parameter, change char_type_node
12779         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12780         two operands.
12781         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12782         the inlining expansion on target where the type of the call has same or
12783         narrower precision than unsigned char.
12785 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12787         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12788         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12789         * common.opt (fsave-optimization-record): New option.
12790         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12791         * doc/invoke.texi (-fsave-optimization-record): New option.
12792         * dumpfile.c: Include "optinfo-emit-json.h".
12793         (struct kv_pair): Move to coretypes.h.
12794         (optgroup_options): Make non-static.
12795         (dump_context::end_scope): Call
12796         optimization_records_maybe_pop_dump_scope.
12797         * dumpfile.h (optgroup_options): New decl.
12798         * json.cc: New file.
12799         * json.h: New file.
12800         * optinfo-emit-json.cc: New file.
12801         * optinfo-emit-json.h: New file.
12802         * optinfo.cc: Include "optinfo-emit-json.h".
12803         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12804         (optinfo_enabled_p): Check optimization_records_enabled_p.
12805         (optinfo_wants_inlining_info_p): Likewise.
12806         * optinfo.h: Update comment.
12807         * profile-count.c (profile_quality_as_string): New function.
12808         * profile-count.h (profile_quality_as_string): New decl.
12809         (profile_count::quality): New accessor.
12810         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12811         and optinfo_emit_json_cc_tests.
12812         * selftest.h (selftest::json_cc_tests): New decl.
12813         (selftest::optinfo_emit_json_cc_tests): New decl.
12814         * toplev.c: Include "optinfo-emit-json.h".
12815         (compile_file): Call optimization_records_finish.
12816         (do_compile): Call optimization_records_start.
12817         * tree-ssa-live.c: Include optinfo.h.
12818         (remove_unused_scope_block_p): Retain inlining information if
12819         optinfo_wants_inlining_info_p returns true.
12821 2018-07-20  Richard Biener  <rguenther@suse.de>
12823         PR debug/86585
12824         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12825         to cover -flto-partition=none.
12827 2018-07-20  Martin Liska  <mliska@suse.cz>
12829         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12830         (get_decl_source_range): Remove unused function.
12832 2018-07-20  Richard Biener  <rguenther@suse.de>
12834         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12835         (struct vn_phi_s): Likewise.
12836         (struct vn_reference_s): Likewise.
12837         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12838         for searching the slot of an entry known to be in the hash itself.
12839         (vn_phi_hasher::equal): Likewise.
12840         (vn_reference_hasher::equal): Likewise.
12841         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12842         globals.
12843         (optimistic_info, current_info): Remove, keeping only valid_info.
12844         (vn_reference_lookup_1): Remove fallback lookup.
12845         (vn_reference_lookup_2): Likewise.
12846         (vn_nary_op_lookup_1): Likewise.
12847         (vn_phi_lookup): Likewise.
12848         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12849         hash element.
12850         (vn_reference_insert): Adjust, chain the inserted hash element
12851         at last_inserted_ref.
12852         (vn_reference_insert_pieces): Likewise.
12853         (visit_reference_op_call): Likewise.
12854         (vn_nary_op_insert_into): Chain the inserted hash element at
12855         last_inserted_nary.
12856         (vn_nary_op_insert_pieces): Adjust.
12857         (vn_nary_op_insert): Likewise.
12858         (vn_nary_op_insert_stmt): Likewise.
12859         (vn_phi_insert): Adjust, chain the inserted hash element at
12860         last_inserted_phi.
12861         (process_scc): Remove clearing and copying the optimistic
12862         table.  Instead remove elements inserted during an optimistic
12863         iteration from the single table we maintain.
12864         (init_scc_vn): Adjust.
12865         (free_scc_vn): Likewise.
12866         (sccvn_dom_walker::record_cond): Likewise.
12867         (sccvn_dom_walker::after_dom_children): Likewise.
12869 2018-07-19  Martin Sebor  <msebor@redhat.com>
12871         PR tree-optimization/84047
12872         PR tree-optimization/83776
12873         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12874         (check_array_bounds): Call it.
12876 2018-07-19  Martin Sebor  <msebor@redhat.com>
12878         * align.h (align_flags): Use member initialization.
12880 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12882         * Makefile.in (OBJS): Add optinfo.o.
12883         * coretypes.h (class symtab_node): New forward decl.
12884         (struct cgraph_node): New forward decl.
12885         (class varpool_node): New forward decl.
12886         * dump-context.h: New file.
12887         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12888         "tree-pass.h".
12889         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12890         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12891         (set_alt_dump_file): Likewise.
12892         (dump_context::~dump_context): New dtor.
12893         (dump_gimple_stmt): Move implementation to...
12894         (dump_context::dump_gimple_stmt): ...this new member function.
12895         Add the stmt to any pending optinfo, creating one if need be.
12896         (dump_gimple_stmt_loc): Move implementation to...
12897         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12898         Start a new optinfo and add the stmt to it.
12899         (dump_gimple_expr): Move implementation to...
12900         (dump_context::dump_gimple_expr): ...this new member function.
12901         Add the stmt to any pending optinfo, creating one if need be.
12902         (dump_gimple_expr_loc): Move implementation to...
12903         (dump_context::dump_gimple_expr_loc): ...this new member function.
12904         Start a new optinfo and add the stmt to it.
12905         (dump_generic_expr): Move implementation to...
12906         (dump_context::dump_generic_expr): ...this new member function.
12907         Add the tree to any pending optinfo, creating one if need be.
12908         (dump_generic_expr_loc): Move implementation to...
12909         (dump_context::dump_generic_expr_loc): ...this new member
12910         function.  Add the tree to any pending optinfo, creating one if
12911         need be.
12912         (dump_printf): Move implementation to...
12913         (dump_context::dump_printf_va): ...this new member function.  Add
12914         the text to any pending optinfo, creating one if need be.
12915         (dump_printf_loc): Move implementation to...
12916         (dump_context::dump_printf_loc_va): ...this new member function.
12917         Start a new optinfo and add the stmt to it.
12918         (dump_dec): Move implementation to...
12919         (dump_context::dump_dec): ...this new member function.  Add the
12920         value to any pending optinfo, creating one if need be.
12921         (dump_context::dump_symtab_node): New member function.
12922         (dump_context::get_scope_depth): New member function.
12923         (dump_context::begin_scope): New member function.
12924         (dump_context::end_scope): New member function.
12925         (dump_context::ensure_pending_optinfo): New member function.
12926         (dump_context::begin_next_optinfo): New member function.
12927         (dump_context::end_any_optinfo): New member function.
12928         (dump_context::s_current): New global.
12929         (dump_context::s_default): New global.
12930         (dump_scope_depth): Delete global.
12931         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12932         (dump_symtab_node): New function.
12933         (get_dump_scope_depth): Reimplement in terms of dump_context.
12934         (dump_begin_scope): Likewise.
12935         (dump_end_scope): Likewise.
12936         (selftest::temp_dump_context::temp_dump_context): New ctor.
12937         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12938         (selftest::verify_item): New function.
12939         (ASSERT_IS_TEXT): New macro.
12940         (ASSERT_IS_TREE): New macro.
12941         (ASSERT_IS_GIMPLE): New macro.
12942         (selftest::test_capture_of_dump_calls): New test.
12943         (selftest::dumpfile_c_tests): Call it.
12944         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12945         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12946         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12947         descriptive comment.
12948         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12949         (dump_node, dump_bb): Move these unrelated decls.
12950         (class dump_manager): Add leading comment.
12951         * optinfo.cc: New file.
12952         * optinfo.h: New file.
12954 2018-07-19  Michael Collison  <michael.collison@arm.com>
12955             Richard Henderson <rth@redhat.com>
12957         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12958         (subti): Handle op1 zero.
12959         (subvti4, usub4ti4): New.
12960         (*sub<GPI>3_compare1_imm): New.
12961         (sub<GPI>3_carryinCV): New.
12962         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12963         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12965 2018-07-19  Michael Collison  <michael.collison@arm.com>
12966             Richard Henderson <rth@redhat.com>
12968         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12969         (addti3): Create simpler code if low part is already known to be 0.
12970         (addvti4, uaddvti4): New.
12971         (*add<GPI>3_compareC_cconly_imm): New.
12972         (*add<GPI>3_compareC_cconly): New.
12973         (*add<GPI>3_compareC_imm): New.
12974         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12975         handle constants within this pattern..
12976         (*add<GPI>3_compareV_cconly_imm): New.
12977         (*add<GPI>3_compareV_cconly): New.
12978         (*add<GPI>3_compareV_imm): New.
12979         (add<GPI>3_compareV): New.
12980         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12981         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12982         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12983         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12984         with 'comparison' operator.
12985         (*add<GPI>3_compareV_cconly_imm): Ditto.
12986         (*add<GPI>3_compareV_cconly): Ditto.
12987         (*add<GPI>3_compareV_imm): Ditto.
12988         (add<GPI>3_compareV): Ditto.
12989         (add<mode>3_carryinC): Ditto.
12990         (*add<mode>3_carryinC_zero): Ditto.
12991         (*add<mode>3_carryinC): Ditto.
12992         (add<mode>3_carryinV): Ditto.
12993         (*add<mode>3_carryinV_zero): Ditto.
12994         (*add<mode>3_carryinV): Ditto.
12996 2018-07-19  Michael Collison  <michael.collison@arm.com>
12997             Richard Henderson <rth@redhat.com>
12999         * config/aarch64/aarch64-modes.def (CC_V): New.
13000         * config/aarch64/aarch64-protos.h
13001         (aarch64_addti_scratch_regs): Declare
13002         (aarch64_subvti_scratch_regs): Declare.
13003         (aarch64_expand_subvti): Declare.
13004         (aarch64_gen_unlikely_cbranch): Declare
13005         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
13006         for signed overflow using CC_Vmode.
13007         (aarch64_get_condition_code_1): Handle CC_Vmode.
13008         (aarch64_gen_unlikely_cbranch): New function.
13009         (aarch64_addti_scratch_regs): New function.
13010         (aarch64_subvti_scratch_regs): New function.
13011         (aarch64_expand_subvti): New function.
13013 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13015         * config/aarch64/aarch64-option-extensions.def: New entry for profile
13016         extension.
13017         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
13018         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
13019         extension.
13021 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13023         PR target/83009
13024         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13025         address check not strict.
13027 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13029         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
13030         Umq with Umn.
13031         (store_pair_lanes<mode>): Likewise.
13032         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
13033         enum value 'ADDR_QUERY_LDP_STP_N'.
13034         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
13035         (aarch64_print_address_internal): Add declaration.
13036         (aarch64_print_ldpstp_address): Remove.
13037         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
13038         (aarch64_print_operand): Change printing of 'y'.
13039         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
13040         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
13041         'true' rather than '1'.
13042         * config/aarch64/constraints.md (Uml): Likewise.
13043         (Uml): Rename to Umn.
13044         (Umq): Remove.
13046 2018-07-19  Richard Biener  <rguenther@suse.de>
13048         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
13049         a trailing array.
13050         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
13051         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
13052         (vn_reference_hasher): Likewise.
13053         (struct vn_tables_s): Remove obstack and alloc-pool members.
13054         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
13055         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
13056         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
13057         (vn_reference_insert_pieces): Likewise.
13058         (alloc_vn_nary_op_noinit): Adjust.
13059         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
13060         (vn_phi_eq): Adjust.
13061         (shared_lookup_phiargs): Remove.
13062         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
13063         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
13064         (visit_reference_op_call): Likewise.
13065         (copy_nary, copy_phi, copy_reference): Remove.
13066         (process_scc): Rewind the obstack when iterating.  Do not
13067         copy the elements to valid_info but just move them from one
13068         hashtable to the other.
13069         (allocate_vn_table): Adjust.
13070         (free_vn_table): Likewise.
13071         (init_scc_vn): Likewise.
13072         (free_scc_vn): Likewise.
13074 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
13076         PR target/86560
13077         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
13078         indirect_return as function type attribute.
13079         (ix86_attribute_table): Change indirect_return to function
13080         type attribute.
13081         * doc/extend.texi: Update indirect_return attribute.
13083 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
13085         * wide-int.h (widest2_int): New.
13086         * gimple-fold.c (arith_overflowed_p): Use it.
13087         * tree.h (widest2_int_cst): New.
13088         * tree-vrp.c (wide_int_binop_overflow): Rename from
13089         vrp_int_const_binop.
13090         Rewrite to work on trees.
13091         (extract_range_from_multiplicative_op_1): Abstract code to...
13092         (wide_int_range_min_max): ...here.
13093         (wide_int_range_cross_product): ...and here.
13094         (extract_range_from_binary_expr_1): Abstract overflow code to...
13095         (wide_int_range_mult_wrapping): ...here.
13096         * tree-vrp.h (wide_int_range_cross_product): New.
13097         (wide_int_range_mult_wrapping): New.
13099 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
13100             Julia Koval  <julia.koval@intel.com>
13102         * config/i386/x86-tune-costs.h (skylake_memcpy,
13103         skylake_memset): Replace rep_prefix with unrolling for size 512.
13105 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
13107         PR middle-end/86544
13108         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
13109         comparision with EQ_EXPR in last stmt.
13111 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13113         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
13114         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
13115         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
13116         previously known as "PowerPC AltiVec Built-in Functions".  Move
13117         some material to new subsubsections "PowerPC AltiVec Built-in
13118         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
13119         ISA 2.07".
13120         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
13121         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
13122         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
13123         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
13125 2018-07-18  Richard Biener  <rguenther@suse.de>
13127         PR tree-optimization/86557
13128         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
13129         EXACT_DIV_EXPR.
13131 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
13133         * config/s390/s390.c (s390_function_profiler): Generate CFI.
13135 2018-07-17  Jeff Law  <law@redhat.com>
13137         * config/arm/arm.c (get_label_padding): Update for recent
13138         changes to label_to_alignment.
13140         PR tree-optimization/86010
13141         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
13143         * config/mips/mips.c (vr4130_align_insns): Update for recent
13144         changes to label_to_alignment.
13146         * config/frv/frv.c (frv_label_align): Update for recent changes
13147         to label_to_alignment.
13149         * config/nios2/nios2.c (nios2_label_align): Update for recent
13150         changes which dropped ALIGN_LABELS_LOG.
13152 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
13154         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
13155         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
13157 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
13159         * config/arc/arc.c (arc_label_align): Use align_labels instead of
13160         deprecated align_labels_log.
13162 2018-07-17  Richard Biener  <rguenther@suse.de>
13164         PR lto/86456
13165         * dwarf2out.c (init_sections_and_labels): Always generate
13166         a debug_line_str_section for early LTO debug.
13167         (dwarf2out_finish): Reset debug_line_str_hash output early.
13168         Bump counter for extra dwarf5 .debug_loc labels to not conflict
13169         with fat LTO part.
13170         (dwarf2out_early_finish): Output debug_line_str.
13172 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
13174         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
13175         index register on z196 or later.
13177 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
13179         * config/s390/s390.c (s390_default_align): Set default function
13180         alignment to 16.
13181         (s390_override_options_after_change): Call s390_default align.
13182         (s390_option_override_internal): Call s390_default_align.
13183         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
13185 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
13187         PR middle-end/86542
13188         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
13189         to _looptemp_ clauses, other than the first two.
13191 2018-07-17  Martin Liska  <mliska@suse.cz>
13193         * opts.c: Do not enable OPT_falign_* for -Os.
13195 2018-07-17  Martin Liska  <mliska@suse.cz>
13197         * align.h (MAX_CODE_ALIGN): New.
13198         (MAX_CODE_ALIGN_VALUE): New.
13199         * common/config/i386/i386-common.c (ix86_handle_option):
13200         (MAX_CODE_ALIGN): Moved to align.h.
13201         * final.c (MAX_CODE_ALIGN): Likewise.
13202         * opts.c (parse_and_check_align_values):
13203         (MAX_CODE_ALIGN): Likewise.
13204         (MAX_CODE_ALIGN_VALUE): Likewise.
13206 2018-07-17  Martin Liska  <mliska@suse.cz>
13208         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
13209         in order to fulfil coding style.
13210         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
13211         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13212         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13213         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
13214         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
13215         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
13216         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
13217         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
13218         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13220 2018-07-17  Martin Liska  <mliska@suse.cz>
13222         * align.h: New file.
13223         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
13224         directly.
13225         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
13226         align_flags of label_to_alignment.
13227         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
13228         align_flags class.
13229         * config/m68k/m68k.c: Do not use removed align_labels_value and
13230         align_loops_value.
13231         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
13232         (LOOP_ALIGN): Likewise.
13233         (LABEL_ALIGN): Likewise.
13234         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
13235         Remove not used macro.
13236         (rs6000_loop_align): Change return type to align_flags.
13237         (rs6000_loop_align_max_skip): Remove.
13238         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
13239         Change return type to align_flags.
13240         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
13241         Remove not used macro.
13242         (rs6000_loop_align):  Change return type to align_flags.
13243         (rs6000_loop_align_max_skip): Remove.
13244         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
13245         * config/rx/rx-protos.h (rx_align_for_label): Make it
13246         static function.
13247         * config/rx/rx.c (rx_align_for_label): Change return type
13248         to align_flags.
13249         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
13250         macro definitions.
13251         into align_flags class.
13252         (LABEL_ALIGN): Likewise.
13253         (LOOP_ALIGN): Likewise.
13254         * config/s390/s390.c (s390_label_align): Use align_flags
13255         class member.
13256         (s390_asm_output_function_label): Likewise.
13257         * config/sh/sh.c (sh_override_options_after_change):
13258         Use align_flags class directly without macros.
13259         (find_barrier): Likewise.
13260         (barrier_align): Likewise.
13261         (sh_loop_align): Likewise.
13262         * config/spu/spu.c (spu_option_override):
13263         Use align_flags_tuple::get_value instead of removed macros.
13264         (spu_sched_init): Likewise.
13265         * config/spu/spu.h (GTY): Likewise.
13266         * config/visium/visium.c (visium_option_override):
13267         Set "8" as default secondary alignment.
13268         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
13269         in order to guarantee secondary alignment of 8.
13270         * coretypes.h: Include align.h header file.
13271         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
13272         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
13273         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
13274         * doc/tm.texi.in: Likewise.
13275         * final.c (struct label_alignment): Remove not used structure.
13276         (LABEL_ALIGN): Change type to align_flags.
13277         (LOOP_ALIGN): Likewise.
13278         (JUMP_ALIGN): Likewise.
13279         (default_loop_align_max_skip): Remove.
13280         (default_label_align_max_skip): Likewise.
13281         (default_jump_align_max_skip): Likewise.
13282         (default_label_align_after_barrier_max_skip):
13283         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
13284         (LABEL_TO_MAX_SKIP): Remove.
13285         (label_to_alignment): Return align_flags type instead of integer.
13286         (label_to_max_skip): Remove.
13287         (align_fuzz): Use align_flags type.
13288         (compute_alignments): Use align_flags type and use align_flags::max
13289         to combine multiple alignments.
13290         (grow_label_align): Grow vec instead of C array.
13291         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
13292         (shorten_branches):  Use align_flags type and use align_flags::max
13293         to combine multiple alignments.
13294         (final_scan_insn_1): Remove usage of secondary alignment that comes
13295         from label alignment, but instead use proper secondary alignment
13296         which is computed in grow_label_align.
13297         * flags.h (struct align_flags_tuple): Move to align.h.
13298         (struct align_flags): Likewise.
13299         (state_align_loops): Rename to align_loops.
13300         (state_align_jumps): Rename to align_jumps.
13301         (state_align_labels): Rename to align_labels.
13302         (state_align_functions): Rename to align_functions.
13303         (align_loops_log): Remove.
13304         (align_jumps_log): Remove.
13305         (align_labels_log): Remove.
13306         (align_functions_log): Remove.
13307         (align_loops_max_skip): Remove.
13308         (align_jumps_max_skip): Remove.
13309         (align_labels_max_skip): Remove.
13310         (align_functions_max_skip): Remove.
13311         (align_loops_value): Remove.
13312         (align_jumps_value): Remove.
13313         (align_labels_value): Remove.
13314         (align_functions_value): Remove.
13315         * output.h (label_to_alignment): Change return type to align_flags.
13316         (label_to_max_skip): Remove.
13317         * target.def: Remove loop_align_max_skip, label_align_max_skip,
13318         jump_align_max_skip macros.
13319         * targhooks.h (default_loop_align_max_skip): Remove.
13320         (default_label_align_max_skip): Likewise.
13321         (default_jump_align_max_skip): Likewise.
13322         (default_label_align_after_barrier_max_skip): Remove.
13323         * toplev.c (read_log_maxskip): Use ::normalize function.
13324         (parse_N_M): Remove not used argument and also call ::normalize.
13325         (parse_alignment_opts): Do not pass unused arguments.
13326         * varasm.c (assemble_start_function): Use directly align_functions
13327         instead of removed macros.
13328         * system.h: Do not poison removed macros.
13330 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
13332         PR middle-end/86539
13333         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
13334         and cond temporaries don't have reference type if iterator has
13335         pointer type.  For init use &for_pre_body instead of pre_p if
13336         for_pre_body is non-empty.
13338 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13340         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
13341         double-double modes to SFmode directly directly.
13342         (trunc<mode>sf2_fprs): Delete.
13344 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13346         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
13347         for conversions between IFmode and the decimal floating point modes.
13348         (init_float128_ieee): Use the correct names for conversions between
13349         KFmode and the decimal floating point modes.
13351 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13353         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
13354         for the conversions between TDmode and IFmode.
13355         (init_float128_ieee): Use more correct names for the conversions
13356         between TDmode and KFmode.
13358 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
13360         PR tree-optimization/86526
13361         * builtins.c (expand_builtin_memcmp): Formatting fixes.
13362         (inline_expand_builtin_string_cmp): Likewise.
13363         (inline_string_cmp): Likewise.  Use c_readstr instead of
13364         builtin_memcpy_read_str.  Add unit_mode temporary.
13366 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13368         PR middle-end/86528
13369         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
13370         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
13372 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13374         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
13375         Alphabetize prototypes of built-in functions, separating out
13376         built-in functions that are listed in this section but should be
13377         described elsewhere.
13379 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
13381         PR target/86511
13382         * expmed.c (emit_store_flag): Do not emit setcc followed by a
13383         conditional move when trapping comparison was split to a
13384         non-trapping one (and vice versa).
13386 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13388         * config/s390/s390.c (s390_function_profiler): Generate nops
13389         instead of profiler call sequences.
13390         * config/s390/s390.opt: Add the new option.
13392 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13394         * config/s390/s390.c (s390_function_profiler): Generate
13395         __mcount_loc section.
13396         * config/s390/s390.opt: Add the new option.
13398 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13400         * common.opt: Add the new warning.
13401         * config/s390/s390.c (s390_function_profiler): Emit "brasl
13402         %r0,__fentry__" when -mfentry is specified.
13403         (s390_option_override_internal): Disallow -mfentry for 31-bit
13404         CPUs.
13405         * config/s390/s390.opt: Add the new option.
13407 2018-07-16  Richard Biener  <rguenther@suse.de>
13409         PR lto/86523
13410         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
13411         for function-local FUNCTION_DECL and RESULT_DECL immediately.
13413 2018-07-16  Martin Liska  <mliska@suse.cz>
13415         PR ipa/86529
13416         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
13417         to ::get_create.
13419 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
13421         * config/arc/arcHS.md: Update ARCHS scheduling rules.
13423 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
13425         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
13426         for ARCHS4x.
13427         * config/arc/arc-cpus.def (hs4x): New cpu.
13428         (hs4xd): Likewise.
13429         * config/arc/arc-tables.opt: Regenerate.
13430         * config/arc/arc.c (arc_sched_issue_rate): New function.
13431         (TARGET_SCHED_ISSUE_RATE): Define.
13432         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
13433         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
13434         fpu_cvt.
13435         (attr tune): Add ARCHS4x tune values.
13436         (attr tune_dspmpy): Define.
13437         (*tst): Correct instruction type.
13438         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
13439         * config/arc/arcHS4x.md: New file.
13440         * config/arc/fpu.md: Update instruction type attributes.
13441         * config/arc/t-multilib: Regenerate.
13443 2018-07-16  Tom de Vries  <tdevries@suse.de>
13445         PR debug/86455
13446         * var-tracking.c (vt_initialize): Fix pre_dec handling.
13448 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13450         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
13451         early clobber.
13453 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
13455         PR tree-optimization/86514
13456         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
13457         conversion to a boolean type from a type with greater precision.
13459 2018-07-16  Tom de Vries  <tdevries@suse.de>
13461         * var-tracking.c (vt_initialize): Print adjusted insn slim if
13462         dump_flags request TDF_SLIM.
13464 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
13466         * fold-const.c (int_const_binop_1): Abstract...
13467         (wide_int_binop): ...wide int code here.
13468         (poly_int_binop): ...poly int code here.
13469         Abstract the rest of int_const_binop_1 into int_const_binop.
13470         * fold-const.h (wide_int_binop): New.
13471         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
13472         Remove useless PLUS/MINUS_EXPR case.
13473         (zero_nonzero_bits_from_vr): Move wide int code...
13474         (zero_nonzero_bits_from_bounds): ...here.
13475         (extract_range_from_binary_expr_1): Move mask optimization code...
13476         (range_easy_mask_min_max): ...here.
13477         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
13478         (range_easy_mask_min_max): New.
13480 2018-07-15  Jeff Law  <law@redhat.com>
13482         PR target/85993
13483         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
13484         block.
13486 2018-07-14  Jim Wilson  <jimw@sifive.com>
13488         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
13490 2018-07-14  Paul Koning  <ni1d@arrl.net>
13492         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
13494 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
13496         * lto-streamer-out.c (copy_function_or_variable): Dump info about
13497         copying section.
13499 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
13500             Steve Munroe  <munroesj52@gmail.com>
13502         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
13503         (_mm_andnot_si128): Likewise.
13504         (_mm_or_si128): Likewise.
13505         (_mm_xor_si128): Likewise.
13507 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
13509         PR middle-end/78809
13510         * builtins.c (expand_builtin_memcmp): Inline the calls first
13511         when result_eq is false.
13512         (expand_builtin_strcmp): Inline the calls first.
13513         (expand_builtin_strncmp): Likewise.
13514         (inline_string_cmp): New routine. Expand a string compare
13515         call by using a sequence of char comparison.
13516         (inline_expand_builtin_string_cmp): New routine. Inline expansion
13517         a call to str(n)cmp/memcmp.
13518         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
13519         New option.
13520         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
13522 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
13524         * config/arm/driver-arm.c: Include arm-native.h.
13525         (host_detect_local_cpu): Use auto-generated data tables.
13526         (vendors, arm_cpu_table): Delete.  Move part information to ...
13527         * config/arm/arm-cpus.in: ... here.
13528         * config/arm/parsecpu.awk (gen_native): New function.
13529         (vendor, part): New CPU fields.
13530         (END): Add support for building the native CPU detection tables.
13531         * config/arm/t-arm (arm-native.h): Add build rule.
13532         (driver-arm.o): Add dependency on arm-native.h.
13534 2018-07-13  Richard Biener  <rguenther@suse.de>
13536         PR middle-end/85974
13537         * match.pd (addr1 - addr2): Allow either of the operand to
13538         have a conversion.
13540 2018-07-13  Tom de Vries  <tdevries@suse.de>
13542         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
13543         in remap_ssa_name.
13545 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
13547         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
13548         arrays instead of numbered variables.
13550 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
13552         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
13553         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
13555 2018-07-13  Richard Biener  <rguenther@suse.de>
13557         PR debug/86452
13558         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
13559         instead of get_context_die.
13561 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
13562             Richard Biener  <rguenther@suse.de>
13564         PR middle-end/86489
13565         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
13566         that the loop latch destination where phi is defined.
13568 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
13570         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
13571         (riscv_expand_epilogue): Add assertion to check interrupt mode.
13572         (riscv_set_current_function): Extract getting interrupt type to new
13573         function.
13574         (riscv_get_interrupt_type): New function.
13575         (riscv_merge_decl_attributes): New function, checking interrupt type is
13576         same.
13577         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
13579 2018-07-12  Paul Koning  <ni1d@arrl.net>
13581         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
13582         directive.
13584 2018-07-12  Paul Koning  <ni1d@arrl.net>
13586         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
13587         zero reference, add doloop_end instead.
13588         * doc/md.texi (decrement_and_branch_until_zero): Remove.
13589         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
13590         detail for doloop_end.
13592 2018-07-12  Martin Sebor  <msebor@redhat.com>
13594         PR c/86453
13595         * attribs.c (decl_attributes): Reject conflicting attributes before
13596         calling attribute handlers.
13598 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13600         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
13601          parameter.
13602         (gcc::dump_manager::get_dump_file_name): likewise.
13603         (dump_begin): Likewise.
13604         * dumpfile.h (dump_begin): Update prototype.
13605         (gcc::dump_manager::get_dump_file_name,
13606         gcc::dump_manager::get_dump_file_name): Update prototype.
13608 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13610         * internal-fn.h (vectorizable_internal_fn_p): New function.
13611         * tree-vect-slp.c (compatible_calls_p): Likewise.
13612         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
13613         to internal functions.
13614         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
13616 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13618         * fold-const.h (inverse_conditions_p): Declare.
13619         * fold-const.c (inverse_conditions_p): New function.
13620         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
13621         that test the inverse condition of a conditional internal function.
13622         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
13623         * internal-fn.c (internal_fn_mask_index): Handle conditional
13624         internal functions.
13625         (vectorized_internal_fn_supported_p): New function.
13626         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
13627         (any_pred_load_store): Replace with...
13628         (need_to_predicate): ...this new variable.
13629         (redundant_ssa_names): New variable.
13630         (ifcvt_can_use_mask_load_store): Move initial checks to...
13631         (ifcvt_can_predicate): ...this new function.  Handle tree codes
13632         for which a conditional internal function exists.
13633         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
13634         instead of ifcvt_can_use_mask_load_store.  Update after variable
13635         name change.
13636         (predicate_load_or_store): New function, split out from
13637         predicate_mem_writes.
13638         (check_redundant_cond_expr): New function.
13639         (value_available_p): Likewise.
13640         (predicate_rhs_code): Likewise.
13641         (predicate_mem_writes): Rename to...
13642         (predicate_statements): ...this.  Use predicate_load_or_store
13643         and predicate_rhs_code.
13644         (combine_blocks, tree_if_conversion): Update after above name changes.
13645         (ifcvt_local_dce): Handle redundant_ssa_names.
13646         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
13647         general conditional functions.
13648         * tree-vect-stmts.c (vectorizable_call): Likewise.
13650 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13651             Alan Hayward  <alan.hayward@arm.com>
13652             David Sherwood  <david.sherwood@arm.com>
13654         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
13655         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
13656         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
13657         plus and minus and convert them into IFN_COND_FMA-based sequences.
13658         (convert_mult_to_fma): Handle conditional plus and minus.
13660 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13662         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
13663         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
13664         (cond_fnms_optab): New optabs.
13665         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
13666         internal functions.
13667         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
13668         * internal-fn.h (get_conditional_internal_fn): Declare.
13669         (get_unconditional_internal_fn): Likewise.
13670         * internal-fn.c (cond_ternary_direct): New macro.
13671         (expand_cond_ternary_optab_fn): Likewise.
13672         (direct_cond_ternary_optab_supported_p): Likewise.
13673         (FOR_EACH_COND_FN_PAIR): Likewise.
13674         (get_conditional_internal_fn): New function.
13675         (get_unconditional_internal_fn): Likewise.
13676         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
13677         (gimple_match_op::gimple_match_op): Add a new overload for 5
13678         operands.
13679         (gimple_match_op::set_op): Likewise.
13680         (gimple_resimplify5): Declare.
13681         * genmatch.c (decision_tree::gen): Generate simplifications for
13682         5 operands.
13683         * gimple-match-head.c (gimple_simplify): Define an overload for
13684         5 operands.  Handle calls with 5 arguments in the top-level overload.
13685         (convert_conditional_op): Handle conversions from unconditional
13686         internal functions to conditional ones.
13687         (gimple_resimplify5): New function.
13688         (build_call_internal): Pass a fifth operand.
13689         (maybe_push_res_to_seq): Likewise.
13690         (try_conditional_simplification): Try converting conditional
13691         internal functions to unconditional internal functions.
13692         Handle 3-operand unconditional forms.
13693         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13694         Define ternary equivalents of the current rules for binary conditional
13695         internal functions.
13696         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13697         ternary operations.
13698         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13699         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13700         (optab): Handle them.
13701         (SVE_COND_FP_TERNARY): New int iterator.
13702         (sve_fmla_op, sve_fmad_op): New int attributes.
13703         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13704         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13705         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13707 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13709         * target.def (preferred_else_value): New target hook.
13710         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13711         * doc/tm.texi: Regenerate.
13712         * targhooks.h (default_preferred_else_value): Declare.
13713         * targhooks.c (default_preferred_else_value): New function.
13714         * internal-fn.h (conditional_internal_fn_code): Declare.
13715         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13716         (get_conditional_internal_fn): Use it.
13717         (conditional_internal_fn_code): New function.
13718         * gimple-match.h (gimple_match_cond): New struct.
13719         (gimple_match_op): Add a cond member function.
13720         (gimple_match_op::gimple_match_op): Update all forms to take a
13721         gimple_match_cond.
13722         * genmatch.c (expr::gen_transform): Use the same condition as res_op
13723         for the suboperation, but don't specify a particular else_value.
13724         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13725         (visit_nary_op, visit_reference_op_load): Pass
13726         gimple_match_cond::UNCOND to the gimple_match_op constructor.
13727         * gimple-match-head.c: Include tree-eh.h
13728         (convert_conditional_op): New function.
13729         (maybe_resimplify_conditional_op): Likewise.
13730         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13731         (gimple_resimplify2): Likewise.
13732         (gimple_resimplify3): Likewise.
13733         (gimple_resimplify4): Likewise.
13734         (maybe_push_res_to_seq): Return null for conditional operations.
13735         (try_conditional_simplification): New function.
13736         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
13737         constructor.
13738         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13739         IFN_COND_* call.
13740         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13741         function.
13742         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13744 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13746         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13747         DECL_FCONTEXT
13748         (hash_tree): Do not hash DECL_FCONTEXT
13749         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13750         Do not stream DECL_FCONTEXT.
13751         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13752         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13754 2018-07-12  Richard Biener  <rguenther@suse.de>
13756         PR debug/86462
13757         * dwarf2out.c (gen_block_die): Only output blocks when they have
13758         at least one !DECL_IGNORED_P variable.
13760 2018-07-12  Richard Biener  <rguenther@suse.de>
13762         PR target/84829
13763         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13764         Remove -mieee-fp handling.
13766 2018-07-12  Richard Biener  <rguenther@suse.de>
13768         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13769         left-over from last patch.
13771 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
13773         PR tree-optimization/86492
13774         * gimple-ssa-store-merging.c
13775         (imm_store_chain_info::coalesce_immediate_stores): Call
13776         check_no_overlap even for the merge_overlapping case.  Formatting fix.
13778 2018-07-12  Richard Biener  <rguenther@suse.de>
13780         PR middle-end/86479
13781         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13782         move possibly trapping operations into the conditional.
13784 2018-07-12  Richard Biener  <rguenther@suse.de>
13786         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13787         (vn_lookup_simplify_result): Remove recursion limit applied
13788         here.
13789         (vn_nary_build_or_lookup_1): Adjust.
13790         (try_to_simplify): Likewise.
13791         * gimple-match-head.c (gimple_resimplify1): Instead apply one
13792         here.
13793         (gimple_resimplify2): Likewise.
13794         (gimple_resimplify3): Likewise.
13795         (gimple_resimplify4): Likewise.
13797 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
13799         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13800         Use __mmask64 type instead of __mmask8 for __M argument.
13801         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13802         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13803         __U argument.
13804         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13805         __mmask16 for __M argument.
13806         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13807         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13808         to __mmask16 instead of __mmask8.
13809         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13810         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13811         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13812         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13813         instead of __mmask16 for __U argument.
13814         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13815         __mmask16 instead of __mmask8 for __U argument.
13816         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13817         __U argument.
13818         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13819         __mmask16.
13820         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13821         argument.
13822         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13823         __U argument.
13824         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13825         __mmask16.
13826         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13827         of __mmask16.
13828         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13829         __U argument.
13830         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13831         __U argument.
13832         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13833         __U argument.
13834         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13835         __U argument.
13836         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13837         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13838         return type as well as __M argument type and all casts from __mmask8
13839         to __mmask32.
13840         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13841         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13842         return type as well as __M argument type and all casts from __mmask8
13843         to __mmask16.
13844         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13845         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13846         return type as well as __M argument type and all casts from __mmask8
13847         to __mmask32.
13848         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13849         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13850         return type as well as __M argument type and all casts from __mmask8
13851         to __mmask16.
13852         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13853         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13854         __mmask16.
13856 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13858         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13859         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13860         for __U argument.
13862 2018-07-11  Paul Koning  <ni1d@arrl.net>
13864         * doc/md.texi (define_subst): Document how multiple occurrences of
13865         the same argument in the replacement pattern are handled.
13867 2018-07-11  Paul Koning  <ni1d@arrl.net>
13869         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13870         alphabetical order.
13871         (Common Type Attributes): Add "mode" attribute.
13873 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13875         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13876         stream DECL_ORIGINAL_TYPE.
13877         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13878         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13879         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13880         Do not walk original type.
13881         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13882         external decls.
13883         (write_ts_decl_non_common_tree_pointers): Do not stream
13884         DECL_ORIGINAL_TYPE
13885         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13886         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13888 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13890         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13891         thread twice from the same starting edge.
13893 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13895         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13896         * gimple.c (gimple_call_nonnull_result_p): ...here...
13897         (gimple_call_nonnull_arg): ...and here.
13898         * gimple.h (gimple_call_nonnull_result_p): New.
13899         (gimple_call_nonnull_arg): New.
13901 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13903         * config/arm/arm-cpus.in: Move information from fpu field of each
13904         cpu definition to the isa field.
13905         * config/arm/parsecpu.awk (fpu): Delete match rule.
13906         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13908 2018-07-11  Richard Biener  <rguenther@suse.de>
13910         PR debug/86457
13911         * dwarf2out.c (init_sections_and_labels): Use
13912         output_asm_line_debug_info consistently.
13913         (dwarf2out_early_finish): Likewise.
13914         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13915         type units.
13917 2018-07-11  Richard Biener  <rguenther@suse.de>
13919         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13920         Rework father_bb setting in a way to avoid propagating constants
13921         multiple times on a loop body.
13923 2018-07-10  Mark Wielaard  <mark@klomp.org>
13925         PR debug/86459
13926         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13928 2018-07-10  Richard Biener  <rguenther@suse.de>
13930         * hash-map.h (hash_map::iterator::operator*): Return
13931         references to key and value.
13933 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13935         PR c++/86443
13936         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13937         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13938         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13939         (gimplify_omp_for): For composite loops, move outer
13940         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13941         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13942         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13943         TREE_LIST for both the original class iterator and the "last" helper
13944         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13945         loop, remember has_decl_expr from outer composite loops for the
13946         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13948 2018-07-09  Martin Sebor  <msebor@redhat.com>
13950         PR middle-end/77357
13951         PR middle-end/86428
13952         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13953         accessing implicitly initialized array elements.
13954         * expr.c (string_constant): Handle string initializers of
13955         character arrays within aggregates.
13956         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13957         Store element offset.  As a special case, handle zero size.
13958         (fold_nonarray_ctor_reference): Same.
13959         (fold_ctor_reference): Add argument.  Store subobject offset.
13960         * gimple-fold.h (fold_ctor_reference): Add argument.
13962 2018-07-09  Paul Koning  <ni1d@arrl.net>
13964         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13965         (pdp11_insn_cost): New function.
13966         (pdp11_md_asm_adjust): New function.
13967         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13968         (pdp11_rtx_costs): Update to match machine better.
13969         (output_addr_const_pdp11): Correct format mismatch warnings.
13970         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13971         * config/pdp11/pdp11.md: General change to add base_cost and/or
13972         length attributes for use by new pdp11_insn_cost function.
13973         (MIN_BRANCH): Correct definition.
13974         (MIN_SOB): Ditto.
13975         (doloop_end): Use standard pattern name for looping pattern.
13976         (doloop_end_nocc): New.
13977         (movsf): Add another constraint alternative.
13978         (zero_extendqihi2): Add constraint alternatives for not in place
13979         extend.
13980         (zero_extendhisi2): Remove.
13981         (shift patterns): Add CC handling variants.
13982         (bswaphi2): New.
13983         (bswapsi2): New.
13984         (rothi3): New.
13985         (define_peephole2): New peephole to recognize mov that sets CC for
13986         subsequent test.
13988 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13990         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13991         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13992         wi::add.
13994 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13996         PR c/86420
13997         * real.c (real_nextafter): Return true if result is denormal.
13999 2018-07-09  Martin Liska  <mliska@suse.cz>
14001         * common.opt: Add back wrongly removed attribute.
14003 2018-07-09  Richard Biener  <rguenther@suse.de>
14005         PR debug/86413
14006         * dwarf2out.c (gen_block_die): For an early generated DIE
14007         always output high/low PC attributes.
14009 2018-07-09  Tom de Vries  <tdevries@suse.de>
14011         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
14012         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
14013         onto VAR_DECL with abstract origin.
14015 2018-07-07  Jim Wilson  <jimw@sifive.com>
14017         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
14019 2018-07-07  Tom de Vries  <tdevries@suse.de>
14021         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
14023 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
14025         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
14026         overflow_type.
14027         (combine_bound): Use wide-int overflow calculation instead of
14028         rolling our own.
14029         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
14030         overflow_type.
14031         * fold-const.c (int_const_binop_2): Same.
14032         (extract_muldiv_1): Same.
14033         (fold_div_compare): Same.
14034         (fold_abs_const): Same.
14035         * match.pd: Same.
14036         * poly-int.h (add): Same.
14037         (sub): Same.
14038         (neg): Same.
14039         (mul): Same.
14040         * predict.c (predict_iv_comparison): Same.
14041         * profile-count.c (slow_safe_scale_64bit): Same.
14042         * simplify-rtx.c (simplify_const_binary_operation): Same.
14043         * tree-chrec.c (tree_fold_binomial): Same.
14044         * tree-data-ref.c (split_constant_offset_1): Same.
14045         * tree-if-conv.c (idx_within_array_bound): Same.
14046         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14047         * tree-ssa-phiopt.c (minmax_replacement): Same.
14048         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
14049         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
14050         * vr-values.c (vr_values::adjust_range_with_scev): Same.
14051         * wide-int.cc (wi::add_large): Same.
14052         (wi::mul_internal): Same.
14053         (wi::sub_large): Same.
14054         (wi::divmod_internal): Same.
14055         * wide-int.h: Change overflow type to overflow_type for neg, add,
14056         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
14057         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
14058         mul_internal, divmod_internal.
14059         (overflow_type): New enum.
14060         (accumulate_overflow): New.
14062 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
14064         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
14065         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
14067 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
14069         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
14070         argument is checked for zero before entering loop, avoid checking again.
14072 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
14074         * gimplify.h (generic_expr_could_trap_p): Set as global function.
14075         * gimplify.h (generic_expr_could_trap_p): Likwise.
14076         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
14078 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
14080         PR tree-optimization/86401
14081         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
14082         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
14083         (fold_bit_and_mask): ... here.  New helper function for match.pd.
14084         * fold-const.h (fold_bit_and_mask): Declare.
14085         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
14087 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
14089         PR target/86324
14090         * target.def (translate_mode_attribute): New hook.
14091         * targhooks.h (default_translate_mode_attribute): Declare.
14092         * targhooks.c (default_translate_mode_attribute): New function.
14093         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
14094         * doc/tm.texi: Regenerate.
14095         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
14096         (rs6000_translate_mode_attribute): New function.
14098 2018-07-06  Paul Koning  <ni1d@arrl.net>
14100         * doc/md.texi (define_split): Document DONE and FAIL.
14101         (define_peephole2): Ditto.
14103 2018-07-05  Jeff Law  <law@redhat.com>
14105         PR tree-optimization/86010
14106         * tree-ssa-dse.c (compute_trims): More aggressively trim at
14107         both the head and tail of mem* and str* calls.
14109 2018-07-05  Jim Wilson  <jimw@sifive.com>
14111         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
14113 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
14115         * config/aarch64/aarch64-simd.md: correct flags text for
14116         MIN_EXPR replacement.
14118 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
14120         * configure: Regenerated.
14122 2018-07-05  Carl Love  <cel@us.ibm.com>
14124         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
14125         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
14126         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
14127         VSX_BUILTIN_DOUBLEL_V4SF.
14129 2018-07-05  Martin Sebor  <msebor@redhat.com>
14131         PR c++/86400
14132         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
14133         than its domain to compute its the upper bound of a char array.
14135 2018-07-05  Nathan Sidwell  <nathan@acm.org>
14137         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
14138         * doc/cpp.texi: Update comment.
14139         * doc/tm.texi: Rebuilt.
14140         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
14141         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
14142         * doc/extend.texi (Backwards Compatibility): Clarify it is system
14143         headers affected by extern "C".
14144         * system.h: Poison NO_IMPLICIT_EXTERN_C.
14145         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
14146         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
14147         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
14148         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
14149         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
14150         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
14151         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
14152         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
14153         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
14154         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
14155         config/sparc/sp64-elf.h, config/spu/spu.h,
14156         config/stormy16/stormy16.h, config/v850/v850.h,
14157         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
14158         define NO_IMPLICIT_EXTERN_C.
14159         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
14161 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14163         PR target/84711
14164         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14165         instead of GET_MODE_SIZE when comparing Units.
14167 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14169         PR target/84711
14170         * rtlanal.c (set_noop_p): Constrain on mode change,
14171         include hard-reg-set.h
14173 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14175         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
14177 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
14179         Revert
14180         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
14182         PR sanitizer/84250
14183         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14184         libasan.
14185         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14187 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
14189         PR sanitizer/84250
14190         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14191         libasan.
14192         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14194 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
14196         PR middle-end/86380
14197         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
14199 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
14201         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
14202         neg_*_op* variables.
14204 2018-07-04  Martin Liska  <mliska@suse.cz>
14206         * tree-switch-conversion.c: Define
14207         max_ratio_for_speed and max_ratio_for_size constants.
14209 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
14210             Martin Liska  <mliska@suse.cz>
14212         PR middle-end/66240
14213         PR target/45996
14214         PR c/84100
14215         * common.opt: Rename align options with 'str_' prefix.
14216         * common/config/i386/i386-common.c (set_malign_value): New
14217         function.
14218         (ix86_handle_option): Use it to set -falign-* options/
14219         * config/aarch64/aarch64-protos.h (struct tune_params): Change
14220         type from int to string.
14221         * config/aarch64/aarch64.c: Update default values from int
14222         to string.
14223         * config/alpha/alpha.c (alpha_override_options_after_change):
14224         Likewise.
14225         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
14226         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14227         max skip conditionally.
14228         * config/i386/freebsd.h (SUBALIGN_LOG): New.
14229         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14230         max skip conditionally.
14231         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14232         max skip conditionally.
14233         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
14234         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14235         max skip conditionally.
14236         * config/i386/i386.c (struct ptt): Change type from int to
14237         string.
14238         (ix86_default_align): Set default values.
14239         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
14240         max skip conditionally.
14241         * config/i386/iamcu.h (SUBALIGN_LOG): New.
14242         (ASM_OUTPUT_MAX_SKIP_ALIGN):
14243         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
14244         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14245         max skip conditionally.
14246         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
14247         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
14248         * config/i386/x86-64.h (SUBALIGN_LOG): New.
14249         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14250         max skip conditionally.
14251         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14252         * config/ia64/ia64.c (ia64_option_override): Set default values
14253         for alignment options.
14254         * config/m68k/m68k.c: Handle new str_align_* options.
14255         * config/mips/mips.c (mips_set_compression_mode): Change
14256         type of constants.
14257         (mips_option_override): Set default values for options.
14258         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
14259         Likewise.
14260         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14261         Likewise.
14262         * config/rx/rx.c (rx_option_override): Likewise.
14263         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
14264         (LABEL_ALIGN): Use align_labels_log.
14265         (LOOP_ALIGN): Use align_loops_align.
14266         * config/s390/s390.c (s390_asm_output_function_label): Use new
14267         macros.
14268         * config/sh/sh.c (sh_override_options_after_change):
14269         Change type of constants.
14270         * config/spu/spu.c (spu_sched_init): Likewise.
14271         * config/sparc/sparc.c (sparc_option_override): Set default
14272         values for options.
14273         * config/visium/visium.c (visium_option_override): Likewise.
14274         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
14275         emit p2align format with last argument if it's not needed.
14276         * doc/invoke.texi: Document extended format of -falign-*.
14277         * final.c: Use align_labels alignment.
14278         * flags.h (struct target_flag_state): Change type to use
14279         align_flags.
14280         (struct align_flags_tuple): New.
14281         (struct align_flags): Likewise.
14282         (align_loops_log): Redefine macro to use new types.
14283         (align_loops_max_skip): Redefine macro to use new types.
14284         (align_jumps_log): Redefine macro to use new types.
14285         (align_jumps_max_skip): Redefine macro to use new types.
14286         (align_labels_log): Redefine macro to use new types.
14287         (align_labels_max_skip): Redefine macro to use new types.
14288         (align_functions_log): Redefine macro to use new types.
14289         (align_loops): Redefine macro to use new types.
14290         (align_jumps): Redefine macro to use new types.
14291         (align_labels): Redefine macro to use new types.
14292         (align_functions): Redefine macro to use new types.
14293         (align_functions_max_skip): Redefine macro to use new types.
14294         (align_loops_value): New macro.
14295         (align_jumps_value): New macro.
14296         (align_labels_value): New macro.
14297         (align_functions_value): New macro.
14298         * function.c (invoke_set_current_function_hook): Propagate
14299         alignment values from flags to global variables default in
14300         topleev.h.
14301         * ipa-icf.c (sem_function::equals_wpa): Use
14302         cl_optimization_option_eq instead of memcmp.
14303         * lto-streamer.h (cl_optimization_stream_out): Support streaming
14304         of string types.
14305         (cl_optimization_stream_in): Likewise.
14306         * optc-save-gen.awk: Support strings in cl_optimization.
14307         * opth-gen.awk: Likewise.
14308         * opts.c (finish_options): Remove error checking of invalid
14309         value ranges.
14310         (MAX_CODE_ALIGN): Remove.
14311         (MAX_CODE_ALIGN_VALUE): Likewise.
14312         (parse_and_check_align_values): New function.
14313         (check_alignment_argument): Likewise.
14314         (common_handle_option): Use check_alignment_argument.
14315         * opts.h (parse_and_check_align_values): Declare.
14316         * toplev.c (init_alignments): Remove.
14317         (read_log_maxskip): New.
14318         (parse_N_M): Likewise.
14319         (parse_alignment_opts): Likewise.
14320         (backend_init_target): Remove usage of init_alignments.
14321         * toplev.h (parse_alignment_opts): Declare.
14322         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
14323         argument.
14324         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
14325         * tree.c (cl_option_hasher::equal): New.
14326         * varasm.c: Use new global macros.
14328 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
14330         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
14331         Use a simpler align directive also if MAXSKIP = ALIGN-1.
14332         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14333         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14334         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14335         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14336         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
14337         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
14338         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
14339         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14340         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14341         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14342         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14344 2018-07-04  Martin Liska  <mliska@suse.cz>
14345             Jonathan Wakely  <jwakely@redhat.com>
14347         * coverage.c: Use correct type.
14348         * doc/invoke.texi: Language correction.
14350 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
14352         PR target/85620
14353         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
14354         ENDBRANCH for non-tail call which may return via indirect branch.
14355         * doc/extend.texi: Document indirect_return attribute.
14357 2018-07-03  Martin Sebor  <msebor@redhat.com>
14359         PR tree-optimization/86274
14360         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
14361         precondition.
14362         (format_floating): Correct handling of infinities and NaNs.
14364 2018-07-03  Martin Sebor  <msebor@redhat.com>
14366         * print-tree.c (print_real_cst): New function.
14367         (print_node_brief): Call it.
14368         (print_node): Ditto.
14370 2018-07-03  Jeff Law  <law@redhat.com>
14372         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
14373         into a single pattern.
14375         * config/h8300/h8300.md (ors code_iterator): New.
14376         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
14377         a single pattern and single splitter.
14378         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
14379         (iorqi3_1, xorqi3_1): Likewise.
14380         (iorqi3, xorqi3 expanders): Similarly.
14382         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
14383         (movmd_internal) into a single pattern using the P mode iterator.
14384         (movmd splitters): Similarly.
14385         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
14386         (movsd splitters): Similarly.
14388         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
14389         ADDB, ADDW and ADDL into a single ADD attribute which selects the
14390         right table based on the size of the operand.
14391         * config/h8300/h8300.md (length_table): Corresponding changes. All
14392         references to "addb", "addw" and "addl" changed to "add".
14393         (btst patterns): Merge two variants into a single pattern.
14394         (tstqi, tsthi): Likewise.
14395         (addhi3_incdec, addsi3_incdec): Likewise.
14396         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
14397         (mulhi3, mulsi3): Likewise.
14398         (udivhi3, udivsi3): Likewise.
14399         (divhi3, divsi3): Likewise.
14400         (andorqi3, andorhi3, andorsi3): Likewise.
14402 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
14404         PR target/85694
14405         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
14406         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
14408 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14410         PR tree-optimization/85694
14411         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
14412         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
14413         UNSPEC_URHADD.
14414         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
14415         (<u>avg<mode>3_ceil): New patterns.
14417 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
14419         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
14420         scan-tree-dump directive.
14421         * gcc.dg/vect/slp-perm-2.c: Likewise.
14422         * gcc.dg/vect/slp-perm-3.c: Likewise.
14423         * gcc.dg/vect/slp-perm-5.c: Likewise.
14424         * gcc.dg/vect/slp-perm-6.c: Likewise.
14425         * gcc.dg/vect/slp-perm-7.c: Likewise.
14426         * gcc.dg/vect/slp-perm-8.c: Likewise.
14428 2018-07-03  Marek Polacek  <polacek@redhat.com>
14430         PR middle-end/86202
14431         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
14433 2018-07-03  Richard Biener  <rguenther@suse.de>
14435         PR ipa/86389
14436         * tree-ssa-structalias.c (find_func_clobbers): Properly
14437         handle indirect calls.
14439 2018-07-03  Jeff Law  <law@redhat.com>
14441         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
14442         (shifts): New code iterator.
14443         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
14444         expander.  Fix HImode handling on H8/SX.
14445         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
14446         (subqi3, subhi3, subsi3 expanders): Likewise.
14447         (andqi3, andhi3, andsi3 expanders): Likewise.
14448         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
14449         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
14450         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
14451         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
14452         (zero_extendqihi2, zero_extendqisi2): Likewise.
14453         (extendqihi2, extendqisi2): Likewise.
14454         (rotlqi3, rotlhi3, rotlsi3): Likewise.
14455         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
14456         (rotlqi3_1, rotlhi3_1): Likewise.
14457         (logicalhi3_sn, logicalsi3_sn): Likewise.
14458         (logicalhi3, logicalsi3): Likewise.
14460 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14462         * tree-vect-patterns.c (vect_recog_rotate_pattern)
14463         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14464         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
14465         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
14466         type to append_pattern_def_seq instead of creating a stmt_vec_info
14467         directly.
14468         (build_mask_conversion): Likewise.  Remove vinfo argument.
14469         (vect_add_conversion_to_patterm): Likewise, renaming to...
14470         (vect_add_conversion_to_pattern): ...this.
14471         (vect_recog_mask_conversion_pattern): Update call to
14472         build_mask_conversion.  Pass the vector type to
14473         append_pattern_def_seq here too.
14474         (vect_recog_gather_scatter_pattern): Update call to
14475         vect_add_conversion_to_pattern.
14477 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14479         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
14480         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14481         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
14482         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
14483         STMT_VINFO_PATTERN_DEF_SEQ to null here.
14484         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
14485         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
14486         append_pattern_def_seq instead of new_pattern_def_seq.
14487         (vect_recog_divmod_pattern): Do both of the above.
14488         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
14489         is null.
14491 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14493         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
14494         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
14495         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
14496         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
14497         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
14498         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14499         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
14500         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14501         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14502         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
14503         parameter with a single stmt_vec_info.
14504         (vect_recog_func_ptr): Likewise.
14505         (vect_recog_gather_scatter_pattern): Likewise, folding in...
14506         (vect_try_gather_scatter_pattern): ...this.
14507         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
14508         the stmt_vec_info of the statement to be matched.  Don't clear
14509         STMT_VINFO_RELATED_STMT.
14510         (vect_pattern_recog): Update call accordingly.
14512 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14514         PR tree-optimization/85694
14515         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
14516         (uavgM3_ceil): Document new optabs.
14517         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
14518         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
14519         functions.
14520         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
14521         (savg_ceil_optab): New optabs.
14522         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
14523         (vect_vect_recog_func_ptrs): Add it.
14524         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
14525         constant directly from the associated lhs.
14527 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14529         * tree-vect-patterns.c (vect_split_statement): New function.
14530         (vect_convert_input): Use it to try to split an existing cast.
14532 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14534         * poly-int.h (print_hex): New function.
14535         * dumpfile.h (dump_dec, dump_hex): Declare.
14536         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
14537         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
14538         min_input_precision, operation_precision and operation_sign.
14539         * tree-vect-patterns.c (vect_get_range_info): New function.
14540         (vect_same_loop_or_bb_p, vect_single_imm_use)
14541         (vect_operation_fits_smaller_type): Delete.
14542         (vect_look_through_possible_promotion): Add an optional
14543         single_use_p parameter.
14544         (vect_recog_over_widening_pattern): Rewrite to use new
14545         stmt_vec_info infomration.  Handle one operation at a time.
14546         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
14547         (vect_truncatable_operation_p, vect_set_operation_type)
14548         (vect_set_min_input_precision): New functions.
14549         (vect_determine_min_output_precision_1): Likewise.
14550         (vect_determine_min_output_precision): Likewise.
14551         (vect_determine_precisions_from_range): Likewise.
14552         (vect_determine_precisions_from_users): Likewise.
14553         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
14554         (vect_vect_recog_func_ptrs): Put over_widening first.
14555         Add cast_forwprop.
14556         (vect_pattern_recog): Call vect_determine_precisions.
14558 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14560         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
14561         statements that have been replaced by further pattern statements.
14562         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
14564 2018-07-03  Richard Biener  <rguenther@suse.de>
14566         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
14567         always set *dt.  Dump vectype in vectype overload.
14568         * dumpfile.h (dump_gimple_expr): New function.
14569         (dump_gimple_expr_loc): Likewise.
14570         * dumpfile.c (dump_gimple_expr): New function.
14571         (dump_gimple_expr_loc): Likewise.
14573 2018-07-02  Jeff Law  <law@redhat.com>
14575         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
14576         the H8/300, H8/300H and H8/S variants into a single pattern.
14577         (movhi_h8300, movqi_h8300hs): Similarly.
14578         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
14579         (QHI mode iterator): New.
14581         * config/h8300/h8300.md: Remove trailing whitespace.
14583 2018-07-02  Jim Wilson  <jimw@sifive.com>
14585         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
14586         instead of emit_insn for interrupt returns.
14587         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
14588         (riscv_sret, riscv_uret): Likewise.
14590 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14592         * pretty-print.c (selftest::test_pp_format): Move save and restore
14593         of quotes to class auto_fix_quotes, and add an instance.
14594         * selftest.c: Include "intl.h".
14595         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
14596         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
14597         * selftest.h (selftest::auto_fix_quotes): New class.
14599 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
14601         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
14602         (aarch64_sve_prepare_conditional_op): Remove.
14603         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
14604         Allow aarch64_simd_reg_or_zero as select operand; remove
14605         the aarch64_sve_prepare_conditional_op call.
14606         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
14607         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
14608         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
14609         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
14610         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
14611         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
14612         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
14613         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
14614         and a splitters to match all of the *_any patterns.
14615         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
14617         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
14618         (SVE_COND_FP_BINARY_REV): Remove.
14619         (sve_int_op_rev, sve_fp_op_rev): New.
14620         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
14621         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
14622         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
14623         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
14624         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
14625         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
14626         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
14627         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
14628         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
14630         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
14631         Remove match_dup 1 from the inner unspec.
14632         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
14634         * config/aarch64/aarch64.md (movprfx): New attr.
14635         (length): Default movprfx to 8.
14636         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
14637         (*madd<SVE_I>, *msub<SVE_I): Likewise.
14638         (*<su>mul<SVE_I>3_highpart): Likewise.
14639         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
14640         (*v<ASHIFT><SVE_I>3): Likewise.
14641         (*<su><MAXMIN><SVE_I>3): Likewise.
14642         (*<su><MAXMIN><SVE_F>3): Likewise.
14643         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
14644         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
14645         (*div<SVE_F>4): Likewise.
14647 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
14649         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
14650         in dump string.
14652 2018-07-02  Richard Biener  <rguenther@suse.de>
14654         PR tree-optimization/86363
14655         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
14656         memset argument refers to a non-variable address.
14658 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14660         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
14661         {PLUS,MINUS}_EXPR code to...
14662         (adjust_symbolic_bound): ...here,
14663         (combine_bound): ...here,
14664         (set_value_range_with_overflow): ...and here.
14666 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14668         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
14669         code...
14670         (extract_range_from_abs_expr): ...here.
14672 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
14674         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
14675         -fno-omit-frame-pointer when not optimizing.
14677 2018-07-02  Martin Liska  <mliska@suse.cz>
14679         PR ipa/86279
14680         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
14681         (propagate_nothrow): Likewise.
14683 2018-07-02  Martin Liska  <mliska@suse.cz>
14685         PR ipa/86323
14686         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14688 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14690         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14691         function in r262149, changing "loc" param from source_location to
14692         const dump_location_t &.
14693         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14694         declaration, as above.
14696 2018-07-01  Paul Koning  <ni1d@arrl.net>
14698         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14699         -munit-asm, -mgnu-asm, -mdec-asm.
14700         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14701         (pdp11_output_labelref): New.
14702         (pdp11_output_def): New.
14703         (pdp11_output_addr_vec_elt): New.
14704         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
14705         %# and %@ format codes.
14706         (pdp11_option_override): New.
14707         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14708         (pdp11_output_ident): New.
14709         (pdp11_asm_named_section): New.
14710         (pdp11_asm_init_sections): New.
14711         (pdp11_file_start): New.
14712         (pdp11_file_end): New.
14713         (output_ascii): Use .ascii/.asciz for -mdec-asm.
14714         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
14715         %o, like %c but octal.
14716         (pdp11_option_override): New.
14717         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14718         -mdec-asm.
14719         (DATA_SECTION_ASM_OP): Ditto.
14720         (READONLY_DATA_SECTION_ASM_OP): New.
14721         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14722         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14723         (ASM_OUTPUT_LABELREF): Ditto.
14724         (ASM_OUTPUT_DEF): Ditto.
14725         (ASM_OUTPUT_EXTERNAL): New.
14726         (ASM_OUTPUT_SOURCE_FILENAME): New.
14727         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14728         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14729         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
14730         %# and %@ format codes.
14731         * config/pdp11/pdp11.opt (mgnu-asm): New.
14732         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14733         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14734         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14736 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
14738         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14739         dereferencing path[] beyond its length.
14740         (debug_path): New.
14741         (debug_all_paths): New.
14742         (rewire_first_differing_edge): New.
14743         (adjust_paths_after_duplication): New.
14744         (duplicate_thread_path): Call adjust_paths_after_duplication.
14745         Add new argument.
14746         (thread_through_all_blocks): Add new argument to
14747         duplicate_thread_path.
14749 2018-06-30  Jim Wilson  <jimw@sifive.com>
14751         * config/riscv/predicates.md (p2m1_shift_operand): New.
14752         (high_mask_shift_operand): New.
14753         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14754         pattern using p2m1_shift_operand.
14755         (lshsi3_zero_extend_3+2): New combiner pattern using
14756         high_mask_shift_operand.
14758 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14760         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14761         split out from...
14762         (vect_recog_rotate_pattern): ...here.
14763         (vect_convert_input): Try to insert casts of invariants in the
14764         preheader.
14765         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14766         preheader to be empty.
14768 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14770         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14771         vector type.  If given, install it in the new statement's
14772         STMT_VINFO_VECTYPE.
14773         (vect_element_precision): New function.
14774         (vect_unpromoted_value): New struct.
14775         (vect_unpromoted_value::vect_unpromoted_value): New function.
14776         (vect_unpromoted_value::set_op): Likewise.
14777         (vect_look_through_possible_promotion): Likewise.
14778         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14779         (vect_widened_op_tree, vect_convert_input): Likewise.
14780         (vect_convert_inputs, vect_convert_output): Likewise.
14781         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14782         to handle the optional cast of the multiplication result and
14783         vect_widened_op_tree to detect the widened multiplication itself.
14784         Do not require the input and output of promotion casts to have
14785         the same sign, but base the signedness of the operation on the
14786         input rather than the result.  If the pattern includes two
14787         promotions, check that those promotions have the same sign.
14788         Do not restrict the MULT_EXPR handling to a double-width result;
14789         handle quadruple-width results and wider.  Use vect_convert_inputs
14790         to convert the inputs to the common type.
14791         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
14792         to handle the optional cast of the ABS result.  Also allow a sign
14793         change or a sign extension between the ABS and MINUS.
14794         Use vect_widened_op_tree to detect the widened subtraction and use
14795         vect_convert_inputs to convert the inputs to the common type.
14796         (vect_handle_widen_op_by_const): Delete.
14797         (vect_recog_widen_op_pattern): New function.
14798         (vect_recog_widen_mult_pattern): Use it.
14799         (vect_recog_widen_shift_pattern): Likewise.
14800         (vect_recog_widen_sum_pattern): Use
14801         vect_look_through_possible_promotion to handle the promoted
14802         PLUS_EXPR operand.
14804 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14806         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14807         the containing gimple_seq *.
14808         * gimple-iterator.h (gsi_for_stmt): Declare it.
14809         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14810         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14811         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14812         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14813         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14814         checks.
14815         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14816         split out from...
14817         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14818         statement being replaced is part of an existing pattern
14819         definition sequence, inserting the new pattern statements before
14820         the original one.
14821         (vect_pattern_recog_1): Don't return a bool.  If the statement
14822         is already part of a pattern, instead apply pattern matching
14823         to the pattern definition statements.  Don't clear the
14824         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14825         (vect_pattern_recog): Don't break after the first match;
14826         continue processing the pattern definition statements instead.
14827         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14829 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14831         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14832         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14833         (vect_recog_widen_sum_pattern): Use it.
14835 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14837         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14838         phi is not a pattern statement and has not been replaced by
14839         a pattern statement.
14840         * tree-vect-patterns.c (type_conversion_p): Don't check
14841         STMT_VINFO_IN_PATTERN_P.
14842         (vect_recog_vector_vector_shift_pattern): Likewise.
14843         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14844         the pattern statement rather than the original statement; check
14845         directly for a WIDEN_MULT_EXPR here.
14846         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14847         vect_is_simple_use to return the pattern statement rather
14848         than the original statement; use is_pattern_stmt_p to check
14849         for such a pattern statement.
14850         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14851         to return the pattern statement rather than the original statement;
14852         don't do the same transformation here.
14853         (vect_is_simple_use): If the defining statement has been replaced
14854         by a pattern statement, return the pattern statement instead.
14855         Remove the corresponding (local) transformation from the vectype
14856         overload.
14858 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14860         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14861         end and default to null.
14862         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14863         (vectorizable_reduction): Update calls accordingly, dropping the
14864         gimple ** argument if the passed-back statement isn't needed.
14865         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14866         (vect_recog_rotate_pattern): Likewise.
14867         (vect_recog_mask_conversion_pattern): Likewise.
14868         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14869         (vect_mask_constant_operand_p): Likewise.
14870         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14871         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14872         (get_group_load_store_type, get_load_store_type): Likewise.
14873         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14874         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14875         (vectorizable_conversion, vectorizable_assignment): Likewise.
14876         (vectorizable_shift, vectorizable_operation): Likewise.
14877         (vectorizable_store, vect_is_simple_cond): Likewise.
14878         (vectorizable_condition, vectorizable_comparison): Likewise.
14879         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14880         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14881         and move it to the end.  Cope with null def_stmt_outs.
14883 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14885         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14887 2018-06-29  Jeff Law  <law@redhat.com>
14889         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14890         displacements for TARGET_V850E2V3 and newer.
14891         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14892         * config/v850/v850.md (sign23byte_load): Remove.
14893         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14894         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14896 2018-06-29  Martin Liska  <mliska@suse.cz>
14898         PR lto/85759
14899         * coverage.c (coverage_init): Mangle full path name.
14900         * doc/invoke.texi: Document the change.
14901         * gcov-io.c (mangle_path): New.
14902         * gcov-io.h (mangle_path): Likewise.
14903         * gcov.c (mangle_name): Use mangle_path for path mangling.
14905 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14907         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14908         if starting source register is not even.
14910 2018-06-29  Martin Liska  <mliska@suse.cz>
14912         PR tree-optimization/86263
14913         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14914         Make edge redirection.
14916 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14918         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14919         (dump_scope_depth): New variable.
14920         (get_dump_scope_depth): New function.
14921         (dump_begin_scope): New function.
14922         (dump_end_scope): New function.
14923         * dumpfile.h (get_dump_scope_depth): New declaration.
14924         (dump_begin_scope): New declaration.
14925         (dump_end_scope): New declaration.
14926         (class auto_dump_scope): New class.
14927         (AUTO_DUMP_SCOPE): New macro.
14928         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14929         AUTO_DUMP_SCOPE.
14931 2018-06-29  Richard Biener  <rguenther@suse.de>
14933         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14934         compute_all_dependences succeeds.
14935         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14936         exceed --param loop-max-datarefs-for-datadeps.
14938 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14940         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14942 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14944         PR target/86348
14945         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14946         alternative 0 in preferred_for_speed attribute.
14948 2018-06-28  Paul Koning  <ni1d@arrl.net>
14950         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14951         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14952         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14953         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14954         for shift insn patterns.
14956 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14958         * cgraph.c (cgraph_node::get_body): Replace assignments to
14959         "dump_file" with calls to set_dump_file.
14960         * dumpfile.c (alt_dump_file): Make static, and group with...
14961         (alt_flags): ...this definition.
14962         (dumps_are_enabled): New variable.
14963         (refresh_dumps_are_enabled): New function.
14964         (set_dump_file): New function.
14965         (set_alt_dump_file): New function.
14966         (gcc::dump_manager::dump_start): Replace assignments to
14967         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14968         set_alt_dump_file.
14969         (gcc::dump_manager::dump_finish): Likewise.
14970         * dumpfile.h (alt_dump_file): Delete decl.
14971         (dumps_are_enabled): New variable decl.
14972         (set_dump_file): New function decl.
14973         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14974         global.
14975         * tree-nested.c (lower_nested_functions): Replace assignments to
14976         "dump_file" with calls to set_dump_file.
14978 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14980         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14981         goto_locus of each outgoing edge of each basic block.
14983 2018-06-28  Richard Biener  <rguenther@suse.de>
14985         * dwarf2out.c (decl_scope_table): Remove.
14986         (push_decl_scope): Likewise.
14987         (pop_decl_scope): Likewise.
14988         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14989         (gen_struct_or_union_type_die): Likewise.
14990         (gen_tagged_type_die): Likewise.
14991         (dwarf2out_init): Do not initialize decl_scope_table.
14992         (dwarf2out_c_finalize): Do not free it.
14994 2018-06-28  Richard Biener  <rguenther@suse.de>
14996         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14997         deciding whether to not re-use a DIE.
14999 2018-06-28  Richard Biener  <rguenther@suse.de>
15001         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
15002         DW_AT_abstract_origin attribute.
15004 2018-06-28  Martin Liska  <mliska@suse.cz>
15006         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15007         Use newly introduced constants.
15008         * tree-switch-conversion.h (struct jump_table_cluster):
15009         Define max_ratio_for_size and max_ratio_for_speed.
15011 2018-06-28  Martin Liska  <mliska@suse.cz>
15013         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15014         Add new checking assert to catch invalid state.
15015         (jump_table_cluster::can_be_handled): Handle single case
15016         clusters.
15017         (jump_table_cluster::is_beneficial): Bail out for such case.
15018         (bit_test_cluster::find_bit_tests):
15019         Add new checking assert to catch invalid state.
15020         (bit_test_cluster::can_be_handled): Handle single case
15021         clusters.
15022         (bit_test_cluster::is_beneficial): Bail out for such case.
15023         (switch_decision_tree::analyze_switch_statement):
15024         Fix comment.
15026 2018-06-28  Martin Liska  <mliska@suse.cz>
15028         * common.opt: Introduce -completion option.
15029         * gcc.c (driver_handle_option): Handle it.
15030         (driver::main): Print completions if completion
15031         is set.
15032         * opt-suggestions.c (option_proposer::get_completions):
15033         New function.
15034         (option_proposer::suggest_completion): Likewise.
15035         (option_proposer::find_param_completions): Likewise.
15036         (verify_autocompletions): Likewise.
15037         (test_completion_valid_options): Likewise.
15038         (test_completion_valid_params): Likewise.
15039         (in_completion_p): Likewise.
15040         (empty_completion_p): Likewise.
15041         (test_completion_partial_match): Likewise.
15042         (test_completion_garbage): Likewise.
15043         (opt_proposer_c_tests): Likewise.
15044         * opt-suggestions.h: Declare new functions.
15045         * opts.c (common_handle_option): Handle OPT__completion_.
15046         * selftest-run-tests.c (selftest::run_tests): Add
15047         opt_proposer_c_tests.
15048         * selftest.c (assert_str_startswith): New.
15049         * selftest.h (assert_str_startswith): Likewise.
15050         (opt_proposer_c_tests): New.
15051         (ASSERT_STR_STARTSWITH): Likewise.
15053 2018-06-28  Martin Liska  <mliska@suse.cz>
15055         * Makefile.in: Add opt-suggestions.o.
15056         * gcc-main.c: Include opt-suggestions.h.
15057         * gcc.c (driver::driver): Likewise.
15058         (driver::~driver): Remove m_option_suggestions.
15059         (driver::build_option_suggestions): Moved to option_proposer.
15060         (driver::suggest_option): Likewise.
15061         (driver::handle_unrecognized_options): Use option_proposer.
15062         * gcc.h (class driver): Add new memver m_option_proposer.
15063         * opt-suggestions.c: New file.
15064         * opt-suggestions.h: New file.
15066 2018-06-28  Martin Liska  <mliska@suse.cz>
15068         * vec.h (class auto_string_vec): New (moved from auto_argvec).
15069         (auto_string_vec::~auto_string_vec): Likewise.
15071 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
15073         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
15074         prevent_decl_creation_for_types fields up and add reset_location field.
15075         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
15076         statement if id->reset_location is true.
15077         (copy_edges_for_bb): Do not set goto_locus on the new edges if
15078         id->reset_location is true.
15079         (copy_phis_for_bb): Force input_location on the arguments if
15080         id->reset_location is true.
15081         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
15082         is set on the function to be inlined.
15084 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
15086         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
15088 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
15090         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
15091         registers for Pmode.
15092         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
15093         hard registers for the clobbered pseudo.
15095 2018-06-27  Paul Koning  <ni1d@arrl.net>
15097         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
15098         mutually exclusive options.
15099         * config/pdp11/constraints.md (h): New constraint.
15100         (O): Update definition to match shift code generation.
15101         (D): New constraint.
15102         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
15103         (CCFP): Remove.
15104         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
15105         function.
15106         (output_jump): Change arguments.
15107         (pdp11_fixed_cc_regs): New function.
15108         (pdp11_cc_mode): Ditto.
15109         (pdp11_expand_shift): Ditto.
15110         (pdp11_assemble_shift): Ditto.
15111         (pdp11_small_shift): Ditto.
15112         (pdp11_branch_cost): Remove.
15113         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
15114         from output.
15115         (pdp11_register_move_cost): Update for CC registers.
15116         (pdp11_rtx_costs): Add case for LSHIFTRT.
15117         (pdp11_output_jump): Add CCNZ mode conditional branches.
15118         (notice_update_cc_on_set): Remove.
15119         (pdp11_cc_mode): New function.
15120         (simple_memory_operand): Correct pre/post decrement case.
15121         (no_side_effect_operand): New function.
15122         (pdp11_regno_reg_class): Add CC_REGS class.
15123         (pdp11_fixed_cc_regs): New function.
15124         (pdp11_small_shift): New function.
15125         (pdp11_expand_shift): New function to expand shift insns.
15126         (pdp11_assemble_shift): New function to output shifts.
15127         (pdp11_branch_cost): Remove.
15128         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
15129         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
15130         (WCHAR_TYPE): Ditto.
15131         (PTRDIFF_TYPE): Ditto.
15132         (ADJUST_INSN_LENGTH): New macro.
15133         (FIXED_REGISTERS): Add CC registers.
15134         (CALL_USED_REGISTERS): Ditto.
15135         (reg_class): Ditto.
15136         (REG_CLASS_NAMES): Ditto.
15137         (REG_CLASS_CONTENTS): Ditto.
15138         (SELECT_CC_MODE): Use new function.
15139         (TARGET_FLAGS_REGNUM): New macro.
15140         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
15141         (cc0_reg_rtx): Remove.
15142         (CC_STATUS_MDEP): Remove.
15143         (CC_STATUS_MDEFP_INIT): Remove.
15144         (CC_IN_FPU): Remove.
15145         (NOTICE_UPDATE_CC): Remove.
15146         (REGISTER_NAMES): Add CC registers.
15147         (BRANCH_COST): Change to constant 1.
15148         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
15149         handling.
15150         * config/pdp11/pdp11.opt (mbcopy): Remove.
15151         (mbcopy-builtin): Remove.
15152         (mbranch-cheap): Remove.
15153         (mbranch-expensive): Remove.
15154         * config/pdp11/predicates.md (expand_shift_operand): Update to
15155         match shift code generation.
15156         (ccnz_operator): New predicate.
15157         * doc/invoke.texi (PDP-11 Options): Remove deleted options
15158         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
15159         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
15160         exclusive options.
15161         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
15162         description of O constraint.
15164 2018-06-27  Jeff Law  <law@redhat.com>
15165             Austin Law  <austinklaw@gmail.com>
15167         * config/v850/v850.md (addsi3_set_flags): New pattern.
15168         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
15169         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
15170         (zero_extendhisi2_v850_set_flags): Likewise.
15171         (zero_extendqisi2_v850_set_flags): Likewise.
15172         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
15173         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
15174         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
15176         * config/v850/v850-protos.h (notice_update_cc): Remove.
15177         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
15178         (v850_print_operand): Handle 'D' and "d".
15179         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
15180         Add handling of arithmetic/logical operations compared against zero.
15181         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
15182         Do not look at v850_compare_op, instead get mode from last argument.
15183         (v850_gen_compare): Remove
15184         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
15185         after reload for prologue insns.
15186         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
15187         patterns.
15188         (construct_save_jarl): Likewise.
15189         (TARGET_FLAGS_REGNUM): Define.
15190         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
15191         (NOTICE_UPDATE_CC): Remove.
15192         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
15193         than cc0.  Conditionalize on reload_completed.
15194         (cmpsi_insn, setfcc_insn): Likewise.
15195         (tst1 splitter): Turn into define_and_split which sets the flags
15196         after reload.
15197         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
15198         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
15199         (cstoresf4, cstoredf4): Clobber the flags.
15200         (cmpsi, cmpsf, cmpdf): Remove expanders.
15201         (setf_insn): Remove pattern.
15202         (addsi3): Turn into define_and_split which clobbers the flags after
15203         reload and a suitable pattern (addsi3_clobber_flags) for use after
15204         reload.
15205         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
15206         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
15207         (ashrsi3, ashrsi3_v850e2): Likewise.
15208         (bins): Clobber the flags.
15209         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
15210         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
15211         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
15212         (call_value_internal_short, call_value_internal_long): Likewise.
15213         (callt_save_interrupt, callt_return_interrupt): Likewise.
15214         (save_interrupt, return_interrupt): Likewise.
15215         (callt_save_all_interrupt, save_all_interrupt): Likewise.
15216         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
15217         (restore_all_interrupt, _restore_all_interrupt): Likewise.
15218         (All FP comparisons): Only allow after reload has completed.
15219         (trfsr): Likewise.
15220         (divh, divhu): Tweak output template.
15221         (branch_z_normal, branch_z_invert): Remove
15222         (branch_nz_normal, branch_nz_invert): Likewise.
15223         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
15225         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
15226         * config/v850/v850.c (notice_update_cc): Remove.
15227         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
15228         (CC_NO_CARRY): Likewise.
15229         (NOTICE_UPDATE_CC): Define to nothing.
15230         * config/v850/v850.md: Remove block comment on cc0 handling
15231         Remove "cc" attribute from all patterns.  Remove cc_status handling
15232         from all patterns.  Minor formatting fixes.
15234 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15236         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
15237         (cortex-a76.cortex-a55): Likewise.
15238         * config/aarch64/aarch64-tune.md: Regenerate.
15239         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
15240         cortex-a76.cortex-a55.
15242 2018-06-27  Jeff Law  <law@redhat.com>
15244         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
15245         (MULTILIB_DIRNAMES): Similarly.
15247 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
15249         * gimple.h (gimple_return_retbnd): Delete.
15250         (gimple_return_set_retbnd): Likewise.
15251         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
15252         gimple_return_set_retbnd.
15253         * gimple-pretty-print.c (dump_gimple_return): Remove call to
15254         gimple_return_retbnd and adjust.
15255         * tree-inline.h (struct copy_body_data): Remove retbnd field.
15256         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
15257         Explicitly return NULL in a couple more cases.  Move assertion
15258         on debug statements and remove unreachable code.
15259         (reset_debug_binding): Do not test id->retbnd.
15260         (expand_call_inline): Do not set it.
15262 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
15264         * configure.ac: Add --disable-gcov option.
15265         * configure: Regenerate.
15266         * Makefile.in: Honour @enable_gcov@.
15267         * doc/install.texi: Document --disable-gcov.
15269 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15271         * config/arm/arm-cpus.in (cortex-a76): New entry.
15272         (cortex-a76.cortex-a55): Likewise.
15273         * config/arm/arm-tables.opt: Regenerate.
15274         * config/arm/arm-tune.md: Likewise.
15275         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
15276         * doc/invoke.texi (ARM Options): Document cortex-a76 and
15277         cortex-a76.cortex-a55.
15279 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
15281         PR target/85769
15282         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
15284 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
15286         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
15287         comment.
15288         (EPILOGUE_USES): Likewise.
15290 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
15292         * tree-inline.c (remap_location): New function extracted from...
15293         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
15294         (copy_phis_for_bb): ...here.  Call remap_location.
15295         (copy_cfg_body): Adjust call to copy_edges_for_bb.
15297 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
15299         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
15300         unaligned vsx for 16B memset.
15302 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
15304         PR target/86285
15305         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
15306         ieee128_float_type_node to long_double_type_node unless
15307         TARGET_LONG_DOUBLE_128 is set.
15309 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
15311         * cfgloop.c (get_loop_location): Convert return type from
15312         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
15313         by implicit construction from rtx_insn *, and using
15314         dump_user_location_t::from_function_decl for the fallback case.
15315         * cfgloop.h (get_loop_location): Convert return type from
15316         location_t to dump_user_location_t.
15317         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
15318         dump_printf_loc to pass in a dump_location_t rather than a
15319         location_t, via the gimple stmt.
15320         * coverage.c (get_coverage_counts): Update calls to
15321         dump_printf_loc to pass in dump_location_t rather than a
15322         location_t.
15323         * doc/optinfo.texi (Dump types): Convert example of
15324         dump_printf_loc from taking "locus" to taking "insn".  Update
15325         description of the "_loc" calls to cover dump_location_t.
15326         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
15327         "selftest.h".
15328         (dump_user_location_t::dump_user_location_t): New constructors,
15329         from gimple *stmt and rtx_insn *.
15330         (dump_user_location_t::from_function_decl): New function.
15331         (dump_loc): Make static.
15332         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
15333         const dump_location_t &.
15334         (dump_generic_expr_loc): Delete.
15335         (dump_printf_loc): Convert param "loc" from location_t to
15336         const dump_location_t &.
15337         (selftest::test_impl_location): New function.
15338         (selftest::dumpfile_c_tests): New function.
15339         * dumpfile.h: Include "profile-count.h".
15340         (class dump_user_location_t): New class.
15341         (struct dump_impl_location_t): New struct.
15342         (class dump_location_t): New class.
15343         (dump_printf_loc): Convert 2nd param from source_location to
15344         const dump_location_t &.
15345         (dump_generic_expr_loc): Delete.
15346         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
15347         const dump_location_t &.
15348         * gimple-fold.c (fold_gimple_assign): Update call to
15349         dump_printf_loc to pass in a dump_location_t rather than a
15350         location_t, via the gimple stmt.
15351         (gimple_fold_call): Likewise.
15352         * gimple-loop-interchange.cc
15353         (loop_cand::analyze_iloop_reduction_var): Update for change to
15354         check_reduction_path.
15355         (tree_loop_interchange::interchange): Update for change to
15356         find_loop_location.
15357         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
15358         change in return-type of find_loop_location.
15359         (graphite_regenerate_ast_isl): Likewise.
15360         * graphite-optimize-isl.c (optimize_isl): Likewise.
15361         * graphite.c (graphite_transform_loops): Likewise.
15362         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
15363         pass in a dump_location_t rather than a location_t, via the
15364         gimple stmt.
15365         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15366         * ipa.c (walk_polymorphic_call_targets): Likewise.
15367         * loop-unroll.c (report_unroll): Convert "locus" param from
15368         location_t to dump_location_t.
15369         (decide_unrolling): Update for change to get_loop_location's
15370         return type.
15371         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
15372         location_t to dump_user_location_t.
15373         (grid_find_single_omp_among_assignments_1): Updates calls to
15374         dump_printf_loc to pass in a dump_location_t rather than a
15375         location_t, via the gimple stmt.
15376         (grid_parallel_clauses_gridifiable): Convert "tloc" from
15377         location_t to dump_location_t.  Updates calls to dump_printf_loc
15378         to pass in a dump_location_t rather than a location_t, via the
15379         gimple stmt.
15380         (grid_inner_loop_gridifiable_p): Likewise.
15381         (grid_dist_follows_simple_pattern): Likewise.
15382         (grid_gfor_follows_tiling_pattern): Likewise.
15383         (grid_target_follows_gridifiable_pattern): Likewise.
15384         (grid_attempt_target_gridification): Convert initialization
15385         of local "grid" from memset to zero-initialization; FIXME: does
15386         this require C++11?  Update call to dump_printf_loc to pass in a
15387         optinfo_location rather than a location_t, via the gimple stmt.
15388         * profile.c (read_profile_edge_counts): Updates call to
15389         dump_printf_loc to pass in a dump_location_t rather than a
15390         location_t
15391         (compute_branch_probabilities): Likewise.
15392         * selftest-run-tests.c (selftest::run_tests): Call
15393         dumpfile_c_tests.
15394         * selftest.h (dumpfile_c_tests): New decl.
15395         * tree-loop-distribution.c (pass_loop_distribution::execute):
15396         Update for change in return type of find_loop_location.
15397         * tree-parloops.c (parallelize_loops): Likewise.
15398         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
15399         "locus" from location_t to dump_user_location_t.
15400         (canonicalize_loop_induction_variables): Likewise.
15401         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
15402         for change in return type of find_loop_location.
15403         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
15404         to dump_printf_loc to pass in a dump_location_t rather than a
15405         location_t, via the stmt.
15406         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
15407         Likewise.
15408         * tree-vect-loop-manip.c (find_loop_location): Convert return
15409         type from source_location to dump_user_location_t.
15410         (vect_do_peeling): Update for above change.
15411         (vect_loop_versioning): Update for change in type of
15412         vect_location.
15413         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
15414         from location_t to dump_user_location_t.
15415         (vect_estimate_min_profitable_iters): Update for change in type
15416         of vect_location.
15417         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
15418         location_t to dump_location_t.
15419         (vect_slp_bb): Update for change in type of vect_location.
15420         * tree-vectorizer.c (vect_location): Convert from source_location
15421         to dump_user_location_t.
15422         (try_vectorize_loop_1): Update for change in vect_location's type.
15423         (vectorize_loops): Likewise.
15424         (increase_alignment): Likewise.
15425         * tree-vectorizer.h (vect_location): Convert from source_location
15426         to dump_user_location_t.
15427         (find_loop_location): Convert return type from source_location to
15428         dump_user_location_t.
15429         (check_reduction_path): Convert 1st param from location_t to
15430         dump_user_location_t.
15431         * value-prof.c (check_counter): Update call to dump_printf_loc to
15432         pass in a dump_user_location_t rather than a location_t; update
15433         call to error_at for change in type of "locus".
15434         (check_ic_target): Update call to dump_printf_loc to
15435         pass in a dump_user_location_t rather than a location_t, via the
15436         call_stmt.
15438 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
15440         * config/s390/s390.h (enum processor_flags): Do not use
15441         default tune parameter when -march was specified.
15443 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
15445         PR target/86314
15446         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
15447         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
15448         operands.
15450 2018-06-26  Richard Biener  <rguenther@suse.de>
15452         PR tree-optimization/86287
15453         PR bootstrap/86316
15454         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
15455         (vect_analyze_loop): Initialize n_stmts.
15457 2018-06-26  Richard Biener  <rguenther@suse.de>
15459         PR middle-end/86271
15460         * fold-const.c (fold_convertible_p): Pointer extension
15461         isn't valid.
15463 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
15465         PR debug/86064
15466         * dwarf2out.c (loc_list_has_views): Adjust comments.
15467         (dw_loc_list): Split single cross-partition range with
15468         nonzero locview.
15470 2018-06-25  Jeff Law  <law@redhat.com>
15472         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
15473         on -mbig-switch by default.
15475         * config/v850/predicates.md (const_float_1_operand): Fix match_code
15476         test.
15477         (const_float_0_operand): Remove unused predicate.
15478         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
15479         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
15480         (recipsf2): New expander.  Original pattern now called
15481         (recipsf2_insn).
15482         (recipdf2, recipdf2_insn): Similarly.
15483         (rsqrtsf2, rsqrtsf2_insn): Similarly
15484         (rsqrtdf2, rsqrtdf2_insn): Similarly
15486 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
15488         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
15489         Simplify logic for FreeBSD (twice).
15491 2018-06-25  Martin Sebor  <msebor@redhat.com>
15493         PR tree-optimization/86204
15494         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
15495         a strnlen result if it's less than the length of the string.
15497 2018-06-25  Martin Sebor  <msebor@redhat.com>
15499         PR tree-optimization/85700
15500         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
15501         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
15502         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
15504 2018-06-25  Martin Sebor  <msebor@redhat.com>
15506         * doc/extend.texi (Zero-length arrays): Update and clarify.
15508 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
15510         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
15511         added IEEE/IBM long double multilib support on PowerPC little
15512         endian Linux systems.
15513         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
15514         (MULTILIB_DEFAULTS): Likewise.
15515         * config/rs6000/rs6000.c (rs6000_option_override_internal):
15516         Likewise.
15517         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
15518         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
15519         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
15521 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
15523         PR middle-end/86311
15524         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
15525         (REORDER_45): Likewise.
15527 2018-06-25  Jeff Law  <law@redhat.com>
15529         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
15530         dividend to 32 bits.  Adjust length.
15531         (udivmodhi4): Cleanup output template.  Fix length.
15533 2018-06-25  Carl Love  <cel@us.ibm.com>
15535         * config/rs6000/vsx.md: Change word selector to prefered location.
15537 2018-06-25  Richard Biener  <rguenther@suse.de>
15539         PR tree-optimization/86304
15540         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
15541         epilogue-if-converted loops as well.
15543 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
15545         * lto-section-out.c (lto_begin_section): Do not print section
15546         name for noaddr and unnumbered dumps.
15548 2018-06-25  Richard Biener  <rguenther@suse.de>
15550         * tree-vectorizer.h (struct vec_info_shared): New structure
15551         with parts split out from struct vec_info and loop_nest from
15552         struct _loop_vec_info.
15553         (struct vec_info): Adjust accordingly.
15554         (struct _loop_vec_info): Likewise.
15555         (LOOP_VINFO_LOOP_NEST): Adjust.
15556         (LOOP_VINFO_DATAREFS): Likewise.
15557         (LOOP_VINFO_DDRS): Likewise.
15558         (struct _bb_vec_info): Likewise.
15559         (BB_VINFO_DATAREFS): Likewise.
15560         (BB_VINFO_DDRS): Likewise.
15561         (struct _stmt_vec_info): Add dr_aux member.
15562         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
15563         (DR_MISALIGNMENT_UNINITIALIZED): New.
15564         (set_dr_misalignment): Adjust.
15565         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
15566         (vect_analyze_loop): Adjust prototype.
15567         (vect_analyze_loop_form): Likewise.
15568         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15569         Compute dependences lazily.
15570         (vect_record_base_alignments): Use shared datarefs/ddrs.
15571         (vect_verify_datarefs_alignment): Likewise.
15572         (vect_analyze_data_refs_alignment): Likewise.
15573         (vect_analyze_data_ref_accesses): Likewise.
15574         (vect_analyze_data_refs): Likewise.
15575         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
15576         constructor parameter for shared part.
15577         (vect_analyze_loop_form): Pass in shared part and adjust.
15578         (vect_analyze_loop_2): Pass in storage for the number of
15579         stmts.  Move loop nest finding to the caller.  Compute
15580         datarefs lazily.
15581         (vect_analyze_loop): Pass in shared part.
15582         (vect_transform_loop): Verify shared datarefs are unchanged.
15583         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
15584         constructor parameter for shared part.
15585         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
15586         (vect_slp_bb): Verify shared datarefs are unchanged before
15587         transform.
15588         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
15589         change.
15590         (new_stmt_vec_info): Initialize DR_AUX misalignment to
15591         DR_MISALIGNMENT_UNINITIALIZED.
15592         * tree-vectorizer.c (vec_info::vec_info): Add constructor
15593         parameter for shared part.
15594         (vec_info::~vec_info): Adjust.
15595         (vec_info_shared::vec_info_shared): New.
15596         (vec_info_shared::~vec_info_shared): Likewise.
15597         (vec_info_shared::save_datarefs): Likewise.
15598         (vec_info_shared::check_datarefs): Likewise.
15599         (try_vectorize_loop_1): Construct shared part live for analyses
15600         of a single loop for multiple vector sizes.
15601         * tree-parloops.c (gather_scalar_reductions): Adjust.
15603 2018-06-25  Richard Biener  <rguenther@suse.de>
15605         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
15606         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
15607         (vect_analyze_data_refs): Remove similar code from here and
15608         simplify accordingly.
15610 2018-06-25  Richard Biener  <rguenther@suse.de>
15612         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
15613         for reverse storage order accesses rather than asserting
15614         they cannot happen here.
15616 2018-06-25  Tom de Vries  <tdevries@suse.de>
15618         PR debug/86257
15619         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
15620         Use data16 instead of .byte for insn prefix.
15622 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
15624         PR C++/86082
15625         * parser.c (make_char_string_pack): Pass this literal chars
15626         through cpp_interpret_string.
15627         (cp_parser_userdef_numeric_literal): Check the result of
15628         make_char_string_pack.
15630 2018-06-24  Maya Rashish  <coypu@sdf.org>
15632         * ginclude/stddef.h: Simplify conditions around avoiding
15633         re-definition of __size_t.
15635 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15637         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
15638         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
15640 2018-06-22  Maya Rashish  <coypu@sdf.org>
15642         * doc/invoke.texi (mno-fancy-math-387): Update for changes
15643         made to OpenBSD and NetBSD through the years.
15645 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15647         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15648         behavior of vec_pack (vector double, vector double) to match
15649         behavior of vec_float2 (vector double, vector double).
15651 2018-06-22  Olivier Hainque  <hainque@adacore.com>
15653         * gimplify.c (gimplify_function_tree): Prevent creation
15654         of a trampoline for the address of the current function
15655         passed to entry/exit instrumentation hooks.
15657 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
15659         PR target/86222
15660         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
15661         correctly.
15663 2018-06-22  Martin Liska  <mliska@suse.cz>
15665         PR tree-optimization/86263
15666         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15667         Bail out if is_enabled is false.
15668         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
15669         New declaration.
15670         (jump_table_cluster::is_enabled): New function.
15672 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15674         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15675         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
15676         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
15677         (lto_input_ts_binfo_tree_pointers): Likewise.
15678         * tree-streamer-out.c (streamer_write_tree_bitfields,
15679         write_ts_binfo_tree_pointers): Likewise.
15680         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
15682 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15684         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15686 2018-06-22  Martin Liska  <mliska@suse.cz>
15688         * symbol-summary.h (get): Make it pure and inline move
15689         functionality from ::get function.
15690         (get): Remove and inline into ::get and ::get_create.
15691         (get_create): Move code from ::get function.
15693 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15695         PR target/85994
15696         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15697         -x assembler-with-cpp.
15699 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15701         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15702         _FILE_OFFSET_BITS=64 for C++.
15704 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
15706         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15707         conversion insn that shows up when pr85657-3.c is compiled using
15708         IEEE 128-bit long double.
15709         (neg<mode>2_internal): Use the correct mode to check whether the
15710         mode is IBM extended.
15711         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15712         multiply and divide external functions from being created more
15713         than once.
15715 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15717         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15718         functions.
15719         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
15720         the edge can be forwarded.
15721         (cfg_layout_merge_blocks): Likewise.
15723 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15725         * except.c (finish_eh_generation): Commit edge insertions only after
15726         the EH edges have been redirected from post-landing to landing pads.
15728 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15730         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15731         create_tmp_var_for to create the FRAME decl.
15732         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15734 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15736         * tree-inline.c (copy_edges_for_bb): Minor tweak.
15737         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15738         debug statement when resetting its value.
15739         (expand_call_inline): Copy the locus of the call onto the assignment
15740         of the return value, if any.  Use local variable in more cases.
15742 2018-06-21  Martin Liska  <mliska@suse.cz>
15744         * ipa-pure-const.c (propagate_nothrow): Use
15745         funct_state_summaries->get.
15746         (dump_malloc_lattice): Likewise.
15747         (propagate_malloc): Likewise.
15749 2018-06-21  Richard Biener  <rguenther@suse.de>
15751         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15752         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15753         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15754         comment.
15755         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15756         BLOCK_ABSTRACT_ORIGIN unconditionally.
15758 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
15760         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15761         deleting it.
15762         * ipa-reference.c (ipa_reference_c_finalize): Delete
15763         ipa_ref_opt_sum_summaries and set it to NULL.
15765 2018-06-21  Tom de Vries  <tdevries@suse.de>
15767         PR tree-optimization/85859
15768         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15769         test with comment from bb_no_side_effects_p.
15771 2018-06-21  Richard Biener  <rguenther@suse.de>
15773         PR tree-optimization/86232
15774         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15775         max for constant niter.
15777 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15779         * config/aarch64/aarch64-simd.md
15780         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15782 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15784         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15785         Make opernads of the unspec commutative.
15787 2018-06-21  Richard Biener  <rguenther@suse.de>
15789         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15790         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15791         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15792         (vect_analyze_data_ref_dependence): Re-order checks to deal with
15793         NULL DR_STEP.
15794         (vect_record_base_alignments): Do not record base alignment
15795         for gathers or scatters.
15796         (vect_compute_data_ref_alignment): Drop return value that is always
15797         true.  Bail out early for gathers or scatters.
15798         (vect_enhance_data_refs_alignment): Bail out early for gathers
15799         or scatters.
15800         (vect_find_same_alignment_drs): Likewise.
15801         (vect_analyze_data_refs_alignment): Remove dead code.
15802         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15803         (vect_analyze_data_refs): For possible gathers or scatters do
15804         not create an alternate DR, just check their possible validity
15805         and mark them.  Adjust DECL_NONALIASED handling to not rely
15806         on DR_BASE_ADDRESS.
15807         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15808         update inits of gathers or scatters.
15809         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15810         Also copy gather/scatter flag to pattern vinfo.
15812 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15814         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15815         behavior of vec_packsu (vector unsigned long long, vector unsigned
15816         long long) to match behavior of vec_packs with same signature.
15818 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15819             Thomas Schwinge <thomas@codesourcery.com>
15820             Cesar Philippidis  <cesar@codesourcery.com>
15822         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15823         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15824         (gimplify_adjust_omp_clauses): Likewise.
15825         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15826         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15827         (gimplify_omp_target_update): Update handling of acc update and
15828         enter/exit data.
15829         * omp-low.c (install_var_field): Remove unused parameter
15830         base_pointers_restrict.
15831         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15832         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15833         FINALIZE}
15834         (omp_target_base_pointers_restrict_p): Delete.
15835         (scan_omp_target): Update call to scan_sharing_clauses.
15836         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15837         FINALIZE}.
15838         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15839         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15840         (convert_local_omp_clauses): Likewise.
15841         * tree-pretty-print.c (dump_omp_clause): Likewise.
15842         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15843         FINALIZE}.
15844         (omp_clause_code_name): Likewise.
15846 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15848         PR debug/86194
15849         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15850         be narrowed.
15852         PR tree-optimization/86231
15853         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15854         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15856 2018-06-20  Tom de Vries  <tdevries@suse.de>
15858         PR tree-optimization/86097
15859         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15860         iv type if signedness of iv type is not the same as that of *nit.
15862 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15864         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15865         EDGE_EH edges, verify they are all EDGE_EH.
15867 2018-06-20  Maya Rashish  <coypu@sdf.org>
15869         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15871 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15873         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15874         * config/aarch64/aarch64.c (xgene1_tunings): Add
15875         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15876         (aarch64_mode_valid_for_sched_fusion_p):
15877         Allow 16-byte modes.
15878         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15879         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15880         128-bit modes.
15881         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15882         New pattern.
15883         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15884         * config/aarch64/iterators.md (VQ2): New mode iterator.
15886 2018-06-20  Martin Liska  <mliska@suse.cz>
15888         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15889         Change default ratio from 10 to 8.
15891 2018-06-20  Martin Liska  <mliska@suse.cz>
15893         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15894         New.
15895         (bit_test_cluster::find_bit_tests): Likewise.
15896         (switch_decision_tree::analyze_switch_statement): Find clusters.
15897         * tree-switch-conversion.h (struct jump_table_cluster): Document
15898         hierarchy.
15900 2018-06-20  Martin Liska  <mliska@suse.cz>
15902         * tree-switch-conversion.c (switch_conversion::collect):
15903         Record m_uniq property.
15904         (switch_conversion::expand): Bail out for special conditions.
15905         (group_cluster::~group_cluster): New.
15906         (group_cluster::group_cluster): Likewise.
15907         (group_cluster::dump): Likewise.
15908         (jump_table_cluster::emit): New.
15909         (switch_decision_tree::fix_phi_operands_for_edges): New.
15910         (struct case_node): Remove struct.
15911         (jump_table_cluster::can_be_handled): New.
15912         (case_values_threshold): Moved to header.
15913         (reset_out_edges_aux): Likewise.
15914         (jump_table_cluster::is_beneficial): New.
15915         (bit_test_cluster::can_be_handled): Likewise.
15916         (add_case_node): Remove.
15917         (bit_test_cluster::is_beneficial): New.
15918         (case_bit_test::cmp): New.
15919         (bit_test_cluster::emit): New.
15920         (expand_switch_as_decision_tree_p): Remove.
15921         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15922         (fix_phi_operands_for_edge): Likewise.
15923         (switch_decision_tree::analyze_switch_statement): New.
15924         (compute_cases_per_edge): Move ...
15925         (switch_decision_tree::compute_cases_per_edge): ... here.
15926         (try_switch_expansion): Likewise.
15927         (switch_decision_tree::try_switch_expansion): Likewise.
15928         (record_phi_operand_mapping): Likewise.
15929         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15930         (emit_case_decision_tree): Likewise.
15931         (switch_decision_tree::emit): Likewise.
15932         (balance_case_nodes): Likewise.
15933         (switch_decision_tree::balance_case_nodes): Likewise.
15934         (dump_case_nodes): Likewise.
15935         (switch_decision_tree::dump_case_nodes): Likewise.
15936         (emit_jump): Likewise.
15937         (switch_decision_tree::emit_jump): Likewise.
15938         (emit_cmp_and_jump_insns): Likewise.
15939         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15940         (emit_case_nodes): Likewise.
15941         (switch_decision_tree::emit_case_nodes): Likewise.
15942         (conditional_probability): Remove.
15943         * tree-switch-conversion.h (enum cluster_type): New.
15944         (PRINT_CASE): New.
15945         (struct cluster): Likewise.
15946         (cluster::cluster): Likewise.
15947         (struct simple_cluster): Likewise.
15948         (simple_cluster::simple_cluster): Likewise.
15949         (struct group_cluster): Likewise.
15950         (struct jump_table_cluster): Likewise.
15951         (struct bit_test_cluster): Likewise.
15952         (struct min_cluster_item): Likewise.
15953         (struct case_tree_node): Likewise.
15954         (case_tree_node::case_tree_node): Likewise.
15955         (jump_table_cluster::case_values_threshold): Likewise.
15956         (struct case_bit_test): Likewise.
15957         (struct switch_decision_tree): Likewise.
15958         (struct switch_conversion): Likewise.
15959         (switch_decision_tree::reset_out_edges_aux): Likewise.
15961 2018-06-20  Martin Liska  <mliska@suse.cz>
15963         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15964         (hoist_edge_and_branch_if_true): Likewise.
15965         (expand_switch_using_bit_tests_p): Likewise.
15966         (struct case_bit_test): Likewise.
15967         (case_bit_test_cmp): Likewise.
15968         (emit_case_bit_tests): Likewise.
15969         (switch_conversion::switch_conversion): New class.
15970         (struct switch_conv_info): Remove old struct.
15971         (collect_switch_conv_info): More to ...
15972         (switch_conversion::collect): ... this.
15973         (check_range): Likewise.
15974         (switch_conversion::check_range): Likewise.
15975         (check_all_empty_except_final): Likewise.
15976         (switch_conversion::check_all_empty_except_final): Likewise.
15977         (check_final_bb): Likewise.
15978         (switch_conversion::check_final_bb): Likewise.
15979         (create_temp_arrays): Likewise.
15980         (switch_conversion::create_temp_arrays): Likewise.
15981         (free_temp_arrays): Likewise.
15982         (gather_default_values): Likewise.
15983         (switch_conversion::gather_default_values): Likewise.
15984         (build_constructors): Likewise.
15985         (switch_conversion::build_constructors): Likewise.
15986         (constructor_contains_same_values_p): Likewise.
15987         (switch_conversion::contains_same_values_p): Likewise.
15988         (array_value_type): Likewise.
15989         (switch_conversion::array_value_type): Likewise.
15990         (build_one_array): Likewise.
15991         (switch_conversion::build_one_array): Likewise.
15992         (build_arrays): Likewise.
15993         (switch_conversion::build_arrays): Likewise.
15994         (gen_def_assigns): Likewise.
15995         (switch_conversion::gen_def_assigns): Likewise.
15996         (prune_bbs): Likewise.
15997         (switch_conversion::prune_bbs): Likewise.
15998         (fix_phi_nodes): Likewise.
15999         (switch_conversion::fix_phi_nodes): Likewise.
16000         (gen_inbound_check): Likewise.
16001         (switch_conversion::gen_inbound_check): Likewise.
16002         (process_switch): Use the newly created class.
16003         (switch_conversion::expand): New.
16004         (switch_conversion::~switch_conversion): New.
16005         * tree-switch-conversion.h: New file.
16007 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16009         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
16010         tree-vect-patterns.c.
16011         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
16012         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
16013         (vect_recog_sad_pattern): Likewise.
16014         (vect_recog_widen_sum_pattern): Likewise.
16015         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
16016         (vect_recog_widen_shift_pattern): Remove the type_in argument.
16017         (vect_recog_rotate_pattern): Likewise.
16018         (vect_recog_mult_pattern): Likewise.
16019         (vect_recog_vector_vector_shift_pattern): Likewise.
16020         (vect_recog_divmod_pattern): Likewise.
16021         (vect_recog_mixed_size_cond_pattern): Likewise.
16022         (vect_recog_bool_pattern): Likewise.
16023         (vect_recog_mask_conversion_pattern): Likewise.
16024         (vect_try_gather_scatter_pattern): Likewise.
16025         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
16026         (vect_recog_over_widening_pattern): Likewise.
16027         (vect_recog_gather_scatter_pattern): Likewise.
16028         (vect_recog_func_ptr): Move from tree-vectorizer.h
16029         (vect_vect_recog_func_ptrs): Move further down the file.
16030         (vect_recog_func): Likewise.  Remove the third argument.
16031         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
16032         (vect_pattern_recog_1): Expect the pattern function to do any
16033         necessary target tests.  Also expect it to provide a vector type.
16034         Remove the type_in handling.
16036 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16038         * tree-vect-patterns.c (vect_pattern_detected): New function.
16039         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
16040         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
16041         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
16042         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
16043         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16044         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16045         (vect_recog_mask_conversion_pattern)
16046         (vect_try_gather_scatter_pattern): Likewise.
16048 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16050         * tree-vect-patterns.c (vect_get_internal_def): New function.
16051         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16052         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
16053         (search_type_for_mask_1): Use it.
16055 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16057         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16058         redundant WIDEN_SUM_EXPR handling.
16059         (vect_recog_sad_pattern): Likewise.
16061 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16063         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16064         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
16065         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
16066         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
16067         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
16069 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16071         * tree-vect-stmts.c (vectorizable_call): Make sure that we
16072         use the stmt_vec_info of the original bb statement for the
16073         new zero assignment, even if the call is part of a pattern.
16075 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16077         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
16078         that the sequence is attached to the original statement rather
16079         than the pattern statement.
16080         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
16081         PATTERN_DEF_SEQ from the original statement rather than
16082         the main pattern statement.
16083         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
16084         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
16085         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
16087 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16089         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
16090         definition statements before the early exit for statements that aren't
16091         live or relevant.
16092         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
16093         split out from...
16094         (vect_transform_loop): ...here.  Process pattern definition
16095         statements without first checking whether the main pattern
16096         statement is live or relevant.
16098 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
16100         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
16101         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
16103 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
16105         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
16106         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
16107         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
16108         (expand_block_compare): Change select_block_compare_mode call.
16109         (expand_strncmp_align_check): Use new functions, fix comment.
16110         (emit_final_str_compare_gpr): New function.
16111         (expand_strn_compare): Refactor and clean up code.
16112         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
16114 2018-06-19  Tony Reix  <tony.reix@atos.com>
16115             Damien Bergamini  <damien.bergamini@atos.com>
16116             David Edelsohn  <dje.gcc@gmail.com>
16118         * collect2.c (static_obj): New variable.
16119         (static_libs): New variable.
16120         (is_in_list): Uncomment declaration.
16121         (main): Track AIX libraries linked statically.
16122         (is_in_list): Uncomment definition.
16123         (scan_prog_file): Don't add AIX shared libraries initializer
16124         to constructor list if linking statically.
16126 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
16128         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
16129         constant.
16130         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
16132 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
16134         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
16135         blocks.
16137 2018-06-19  Martin Liska  <mliska@suse.cz>
16139         * config/i386/i386.c (ix86_can_inline_p): Do not use
16140         ipa_fn_summaries::get_create.
16141         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
16142         get.
16143         (devirtualization_time_bonus): Likewise.
16144         (ipcp_propagate_stage): Likewise.
16145         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16146         (edge_set_predicate): Likewise.
16147         (evaluate_conditions_for_known_args): Likewise.
16148         (evaluate_properties_for_edge): Likewise.
16149         (ipa_call_summary::reset): Tranform to ...
16150         (ipa_call_summary::~ipa_call_summary): ... this.
16151         (ipa_fn_summary::reset): Transform to ...
16152         (ipa_fn_summary::~ipa_fn_summary): ... this.
16153         (ipa_fn_summary_t::remove): Rename to ...
16154         (ipa_fn_summary_t::remove_callees): ... this.
16155         (ipa_fn_summary_t::duplicate): Use placement new
16156         instead of memory copy.
16157         (ipa_call_summary_t::duplicate): Likewise.
16158         (ipa_call_summary_t::remove): Remove.
16159         (dump_ipa_call_summary): Change get_create to get.
16160         (ipa_dump_fn_summary): Dump only when summary exists.
16161         (analyze_function_body): Use symbol_summary::get instead
16162         of get_create.
16163         (compute_fn_summary): Likewise.
16164         (estimate_edge_devirt_benefit): Likewise.
16165         (estimate_edge_size_and_time): Likewise.
16166         (inline_update_callee_summaries): Likewise.
16167         (remap_edge_change_prob): Likewise.
16168         (remap_edge_summaries): Likewise.
16169         (ipa_merge_fn_summary_after_inlining): Likewise.
16170         (write_ipa_call_summary): Likewise.
16171         (ipa_fn_summary_write): Likewise.
16172         (ipa_free_fn_summary): Likewise.
16173         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
16174         (struct ipa_call_summary): Likewise.
16175         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
16176         of get_create.
16177         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
16178         (estimate_size_after_inlining): Likewise.
16179         (estimate_growth): Likewise.
16180         (growth_likely_positive): Likewise.
16181         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16182         (inline_call): Likewise.
16183         * ipa-inline.c (caller_growth_limits): Likewise.
16184         (can_inline_edge_p): Likewise.
16185         (can_inline_edge_by_limits_p): Likewise.
16186         (compute_uninlined_call_time): Likewise.
16187         (compute_inlined_call_time): Likewise.
16188         (want_inline_small_function_p): Likewise.
16189         (edge_badness): Likewise.
16190         (update_caller_keys): Likewise.
16191         (update_callee_keys): Likewise.
16192         (inline_small_functions): Likewise.
16193         (inline_to_all_callers_1): Likewise.
16194         (dump_overall_stats): Likewise.
16195         (early_inline_small_functions): Likewise.
16196         (early_inliner): Likewise.
16197         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16198         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16199         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16200         * ipa-split.c (execute_split_functions): Likewise.
16201         * symbol-summary.h: Likewise.
16202         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16204 2018-06-19  Richard Biener  <rguenther@suse.de>
16206         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
16207         (vectorize_loops): ... here.  Fix dbgcnt handling.
16208         (try_vectorize_loop): Wrap try_vectorize_loop_1.
16210 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
16212         PR target/86197
16213         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
16214         ieee128 argument takes up only one (vector) register, not two (floating
16215         point) registers.
16217 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
16219         * gimplify.c (gimplify_init_constructor): Really never clear for an
16220         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
16222 2018-06-19  Richard Biener  <rguenther@suse.de>
16224         PR tree-optimization/86179
16225         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
16226         after failed recognition.
16228 2018-06-18  Martin Sebor  <msebor@redhat.com>
16230         PR middle-end/85602
16231         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
16232         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
16233         Handle integer subtraction.
16234         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
16235         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
16237 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
16239         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
16240         param from rtx to rtx_insn *.
16241         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
16242         param.
16243         (frv_ifcvt_modify_insn): Likwise.
16244         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
16245         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
16246         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
16247         as_a <rtx_insn *> cast to local "unprotected_region" once
16248         it's been established that it's not NULL or pc_rtx.
16249         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
16250         param "sethi" from rtx to rtx_insn *.
16251         (nds32_group_float_insns): Likewise for param "insn".
16252         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
16253         param.
16254         (vax_output_int_subtract): Likewise.
16255         * config/vax/vax.c (vax_output_int_add): Likewise for param
16256         "insn".
16257         (vax_output_int_subtract): Likewise.
16258         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
16259         (emit_pattern_after): Likewise for param "after".
16260         (emit_insn_after): Likewise.
16261         (emit_jump_insn_after): Likewise.
16262         (emit_call_insn_after): Likewise.
16263         (emit_debug_insn_after): Likewise.
16264         (emit_pattern_before): Likewise for param "before".
16265         (emit_insn_before): Likewise.
16266         (emit_jump_insn_before): Likewise.
16267         * final.c (get_insn_template): Likewise for param "insn", removing
16268         a cast.
16269         * output.h (get_insn_template): Likewise for 2nd param.
16270         * rtl.h (emit_insn_before): Likewise.
16271         (emit_jump_insn_before): Likewise.
16272         (emit_debug_insn_before_noloc): Likewise.
16273         (emit_insn_after): Likewise.
16274         (emit_jump_insn_after): Likewise.
16275         (emit_call_insn_after): Likewise.
16276         (emit_debug_insn_after): Likewise.
16277         (set_insn_deleted): Likewise for param.
16279 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
16281         PR target/85358
16282         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
16283         floating point modes, so that IFmode is numerically greater than
16284         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
16285         to declare the ordering.  This prevents IFmode from being
16286         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
16287         machine.  Include rs6000-modes.h to share the fractional values
16288         between genmodes* and the rest of the compiler.
16289         (IFmode): Likewise.
16290         (KFmode): Likewise.
16291         (TFmode): Likewise.
16292         * config/rs6000/rs6000-modes.h: New file.
16293         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
16294         meaning of rs6000_long_double_size so that 126..128 selects an
16295         appropriate 128-bit floating point type.
16296         (rs6000_option_override_internal): Likewise.
16297         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
16298         (TARGET_LONG_DOUBLE_128): Change the meaning of
16299         rs6000_long_double_size so that 126..128 selects an appropriate
16300         128-bit floating point type.
16301         (LONG_DOUBLE_TYPE_SIZE): Update comment.
16302         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
16303         source and destination to match the standard usage.
16304         (truncifkf2): Likewise.
16305         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
16306         ISA 2.07 to use an explicit clobber, instead of passing in a
16307         temporary.
16308         (copysign<mode>3_soft): Likewise.
16310 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
16312         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
16313         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
16314         (vect_slp_analyze_instance_dependence): Likewise.
16315         (vect_enhance_data_refs_alignment): Likewise.
16316         (vect_analyze_data_refs_alignment): Likewise.
16317         (vect_slp_analyze_and_verify_instance_alignment
16318         (vect_analyze_data_ref_accesses): Likewise.
16319         (vect_prune_runtime_alias_test_list): Likewise.
16320         (vect_analyze_data_refs): Likewise.
16321         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
16322         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16323         (vect_analyze_scalar_cycles_1): Likewise.
16324         (vect_get_loop_niters): Likewise.
16325         (vect_analyze_loop_form_1): Likewise.
16326         (vect_update_vf_for_slp): Likewise.
16327         (vect_analyze_loop_operations): Likewise.
16328         (vect_analyze_loop): Likewise.
16329         (vectorizable_induction): Likewise.
16330         (vect_transform_loop): Likewise.
16331         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
16332         * tree-vect-slp.c (vect_analyze_slp): Likewise.
16333         (vect_make_slp_decision): Likewise.
16334         (vect_detect_hybrid_slp): Likewise.
16335         (vect_slp_analyze_operations): Likewise.
16336         (vect_slp_bb): Likewise.
16337         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
16338         (vectorizable_bswap): Likewise.
16339         (vectorizable_call): Likewise.
16340         (vectorizable_simd_clone_call): Likewise.
16341         (vectorizable_conversion): Likewise.
16342         (vectorizable_assignment): Likewise.
16343         (vectorizable_shift): Likewise.
16344         (vectorizable_operation): Likewise.
16345         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
16347 2018-06-18  Martin Sebor  <msebor@redhat.com>
16349         PR tree-optimization/81384
16350         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
16351         * builtins.c (expand_builtin_strnlen): New function.
16352         (expand_builtin): Call it.
16353         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
16354         * builtins.def (BUILT_IN_STRNLEN): New.
16355         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
16356         Warn for bounds in excess of maximum object size.
16357         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
16358         single-value ranges.  Handle strnlen.
16359         (handle_builtin_strlen): Handle strnlen.
16360         (strlen_check_and_optimize_stmt): Same.
16361         * doc/extend.texi (Other Builtins): Document strnlen.
16363 2018-06-18  Maya Rashish  <coypu@sdf.org>
16365         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
16366         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
16367         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
16369         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
16370         here to ...
16371         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
16373 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16375         * tree.c (escaped_string::escape): Replace cast to char * by
16376         const_cast<char *> (unescaped).
16378 2018-06-18  Nick Clifton  <nickc@redhat.com>
16380         PR 84195
16381         * tree.c (escaped_string): New class.  Converts an unescaped
16382         string into its escaped equivalent.
16383         (warn_deprecated_use): Use the new class to convert the
16384         deprecation message, if present.
16385         (test_escaped_strings): New self test.
16386         (test_c_tests): Add test_escaped_strings.
16387         * doc/extend.texi (deprecated): Add a note that the
16388         deprecation message is affected by the -fmessage-length
16389         option, and that control characters will be escaped.
16390         (#pragma GCC error): Document this pragma.
16391         (#pragma GCC warning): Likewise.
16392         * doc/invoke.texi (-fmessage-length): Document this option's
16393         effect on the #warning and #error preprocessor directives and
16394         the deprecated attribute.
16396 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
16398         * tree.c (decl_value_expr_lookup): Revert latest change.
16399         (decl_value_expr_insert): Likewise.
16401 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
16403         * gimplify.c (nonlocal_vlas): Delete.
16404         (nonlocal_vla_vars): Likewise.
16405         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
16406         referenced VLAs.
16407         (gimplify_body): Do not create and destroy nonlocal_vlas.
16408         * tree-nested.c: Include diagnostic.h.
16409         (use_pointer_in_frame): Tweak.
16410         (lookup_field_for_decl): Add assertion and declare the transformation.
16411         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
16412         internal error when the reference is in a wrong context.  Do not
16413         create a debug decl by default.
16414         (note_nonlocal_block_vlas): Delete.
16415         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
16416         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
16417         create a debug decl by default.
16418         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
16419         call is in a wrong context.
16420         (fixup_vla_decls): New function.
16421         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
16422         debug variables were created.
16423         * tree.c (decl_value_expr_lookup): Add checking assertion.
16424         (decl_value_expr_insert): Likewise.
16426 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
16428         PR middle-end/82479
16429         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
16430         * tree-scalar-evolution.c (interpret_expr): Likewise.
16431         (expression_expensive_p): Likewise.
16432         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
16433         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
16434         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
16435         (ssa_defined_by_minus_one_stmt_p): New.
16437 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
16439         PR middle-end/64946
16440         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
16441         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16442         * dojump.c (do_jump): Likewise.
16443         * expr.c (expand_expr_real_2): Check operand type's sign.
16444         * fold-const.c (const_unop): Handle ABSU_EXPR.
16445         (fold_abs_const): Likewise.
16446         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
16447         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
16448         (strip_sign_op_1): Likesise.
16449         * match.pd: Add new pattern to generate ABSU_EXPR.
16450         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
16451         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16452         * tree-eh.c (operation_could_trap_helper_p): Likewise.
16453         * tree-inline.c (estimate_operator_cost): Likewise.
16454         * tree-pretty-print.c (dump_generic_node): Likewise.
16455         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
16456         * tree.def (ABSU_EXPR): New.
16458 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
16460         PR middle-end/86095
16461         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
16462         documented as preserved for backward compatibility only.
16463         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
16465         PR rtl-optimization/86108
16466         * bb-reorder.c (create_forwarder_block): Renamed to ...
16467         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
16468         jump from new landing pad to the second part.
16469         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
16470         Adjust callers.
16472 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
16474         PR middle-end/85878
16475         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
16476         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
16477         Only call store_expr for halves if the mode is the same.
16479         PR middle-end/86123
16480         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
16481         Fix up comment formatting.
16483 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16485         * typed-splay-tree.h (typed_splay_tree::remove): New function.
16486         (typed_splay_tree::closure,
16487         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
16488         (typed_splay_tree::typed_splay_tree,
16489         typed_splay_tree::operator =): Declared private.
16490         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
16491         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
16492         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
16493         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
16494         typed_splay_tree::splay_tree_splay,
16495         typed_splay_tree::splay_tree_foreach_helper,
16496         typed_splay_tree::splay_tree_insert,
16497         typed_splay_tree::splay_tree_remove,
16498         typed_splay_tree::splay_tree_lookup,
16499         typed_splay_tree::splay_tree_predecessor,
16500         typed_splay_tree::splay_tree_successor,
16501         typed_splay_tree::splay_tree_min,
16502         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
16503         (typed_splay_tree::root, typed_splay_tree::comp,
16504         typed_splay_tree::delete_key,
16505         typed_splay_tree::delete_value): New data members.
16506         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
16507         typed_splay_tree::remove.
16509 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
16511         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
16512         -mginv and -mno-ginv to the assembler.
16513         * config/mips/mips.opt (-mcrc): New option.
16514         (-mginv): Likewise.
16515         * doc/invoke.text (-mcrc): Document.
16516         (-mginv): Likewise.
16518 2018-06-15  Nick Clifton  <nickc@redhat.com>
16520         PR 84195
16521         * tree.c (escaped_string): New class.  Converts an unescaped
16522         string into its escaped equivalent.
16523         (warn_deprecated_use): Use the new class to convert the
16524         deprecation message, if present.
16525         (test_escaped_strings): New self test.
16526         (test_c_tests): Add test_escaped_strings.
16527         * doc/extend.texi (deprecated): Add a note that the
16528         deprecation message is affected by the -fmessage-length
16529         option, and that control characters will be escaped.
16530         (#pragma GCC error): Document this pragma.
16531         (#pragma GCC warning): Likewise.
16532         * doc/invoke.texi (-fmessage-length): Document this option's
16533         effect on the #warning and #error preprocessor directives and
16534         the deprecated attribute.
16536 2018-06-15  Richard Biener  <rguenther@suse.de>
16538         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
16539         here, also noting vector size used.
16540         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
16541         size used in MSG_OPTIMIZED_LOCATIONS dump.
16542         (pass_slp_vectorize::execute): Adjust.
16544 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
16546         PR target/85968
16547         * config/arc/arc.c (arc_return_address_register): Fix
16548         if-condition.
16550 2018-06-15  Richard Biener  <rguenther@suse.de>
16552         PR middle-end/86159
16553         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
16554         leave useless conversion stripping to force_gimple_operand_gsi.
16555         (gimplify_build2): Likewise.
16556         (gimplify_build1): Likewise.
16558 2018-06-15  Richard Biener  <rguenther@suse.de>
16560         PR middle-end/86076
16561         * tree-cfg.c (move_stmt_op): unshare invariant addresses
16562         before adjusting their block.
16564 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
16566         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
16567         multilibs for *-*-rtems*.
16568         * config/riscv/t-rtems: New file.
16570 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
16572         PR middle-end/86122
16573         * match.pd ((A +- CST1) +- CST2): Punt if last resort
16574         unsigned_type_for returns NULL.
16576         PR target/85945
16577         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
16578         subregs of multi-word pseudos unless the float mode has word size.
16580 2018-06-14  Richard Biener  <rguenther@suse.de>
16582         PR middle-end/86139
16583         * tree-vect-generic.c (build_word_mode_vector_type): Remove
16584         duplicate and harmful type_hash_canon.
16585         * tree.c (type_hash_canon): Assert we didn't find ourselves.
16587 2018-06-14  Richard Biener  <rguenther@suse.de>
16589         PR ipa/86124
16590         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
16591         NULL cgraph_node.
16593 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
16595         * config/rtems.h (STDINT_LONG32): Define.
16597 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
16598             Prachi Godbole  <prachi.godbole@imgtec.com>
16600         * config/mips/mips-cpus.def: Define P6600.
16601         * config/mips/mips-tables.opt: Regenerate.
16602         * config/mips/mips.c (mips_ucbranch_type): New enum.
16603         (mips_rtx_cost_data): Add support for P6600.
16604         (mips_issue_rate): Likewise.
16605         (mips_multipass_dfa_lookahead): Likewise.
16606         (mips_avoid_hazard): Likewise.
16607         (mips_reorg_process_insns): Likewise.
16608         (mips_classify_branch_p6600): New function.
16609         * config/mips/mips.h (TUNE_P6600): New define.
16610         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
16611         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
16612         * config/mips/mips.md: Include p6600.md.
16613         (processor): Add p6600.
16614         * config/mips/p6600.md: New file.
16615         * doc/invoke.texi: Add p6600 to supported architectures.
16617 2018-06-13  Martin Sebor  <msebor@redhat.com>
16619         PR tree-optimization/86114
16620         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
16621         of integer types.
16622         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
16624 2018-06-13  Richard Biener  <rguenther@suse.de>
16626         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
16627         Properly set vector type of the intermediate stmt.
16628         * tree-vect-stmts.c (vectorizable_operation): The destination
16629         var always has vectype_out type.
16631 2018-06-13  Jeff Law  <law@redhat.com>
16633         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
16634         integer 0 for argument to print_rtl_with_bb.
16635         (rl78_reorg): Likewise.
16637 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
16639         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
16640         from rtx to rtx_insn *.
16641         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
16642         "label".
16643         (add_sched_insns_for_speculation): Likewise for local "target",
16644         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16645         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
16646         from rtx_insn ** to rtx_code_label **.
16647         (reorg_emit_nops): Likewise.
16648         (c6x_reorg): Likewise for local "call_labels".
16649         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
16650         rtx to rtx_insn *.
16651         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
16652         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
16653         the loops over LABEL_REFS.
16654         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
16655         braf_label.
16656         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16657         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
16658         (split_branches): Strengthen local "olabel" from rtx to
16659         rtx_insn *, adding a safe_as_a cast.
16660         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
16661         to "rtx_insn *".
16662         (add_insn_after): Likewise for first two params.
16663         (add_insn_before): Likewise.
16664         (remove_insn): Likewise for param.
16665         (emit_pattern_before_noloc): Likewise for second and third params.
16666         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
16667         (emit_call_insn_before_noloc): Likewise.
16668         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
16669         to "rtx_insn *".
16670         (emit_barrier_before): Likewise.
16671         (emit_label_before): Strengthen "label" param from "rtx" to
16672         "rtx_code_label *".  Strengthen "before" param from "rtx" to
16673         "rtx_insn *".
16674         (emit_insn_after_1): Strengthen "after" param from "rtx" to
16675         "rtx_insn *".
16676         (emit_pattern_after_noloc): Likewise.
16677         (emit_insn_after_noloc): Likewise.
16678         (emit_jump_insn_after_noloc): Likewise.
16679         (emit_call_insn_after_noloc): Likewise.
16680         (emit_debug_insn_after_noloc): Likewise.
16681         (emit_barrier_after): Likewise.
16682         (emit_label_after): Likewise for both params.
16683         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
16684         "loc" param from "int" to "location_t".
16685         (emit_insn_after_setloc): Likewise.
16686         (emit_jump_insn_after_setloc): Likewise.
16687         (emit_call_insn_after_setloc): Likewise.
16688         (emit_debug_insn_after_setloc): Likewise.
16689         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
16690         "loc" param from "int" to "location_t".
16691         (emit_pattern_before): Convert NULL_RTX to NULL.
16692         (emit_insn_before_setloc): Convert "loc" param from "int" to
16693         "location_t".
16694         (emit_jump_insn_before_setloc): Likewise.
16695         (emit_call_insn_before_setloc): Likewise.
16696         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16697         rtx_insn *.  Convert "loc" param from "int" to "location_t".
16698         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16699         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16700         Convert 3rd param from "int" to "location_t".
16701         (emit_barrier_before, emit_barrier_after, next_real_insn):
16702         Strengthen param from rtx to rtx_insn *.
16703         (emit_label_before): Strengthen 1st param from "rtx" to
16704         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
16705         "rtx_insn *".
16706         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16707         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16708         Strengthen 2nd param from "rtx" to "rtx_insn *".
16709         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16710         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16711         Likewise. Convert 3rd param from "int" to "location_t".
16712         (emit_label_after): Strengthen 1st param from "rtx" to
16713         "rtx_code_label *".
16714         (next_real_insn, remove_insn): Strengthen param from "rtx" to
16715         "rtx_insn *".
16716         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16717         from "rtx" to "rtx_insn *".
16719 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
16721         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16722         bodies streamed in with -Q.
16723         * dumpfile.c (dump_files): Add lto-stream-out dump file.
16724         * dumpfile.h (tree_dump_index): Add lto_stream_out.
16725         * gimple-streamer-out.c: Include gimple-pretty-print.h
16726         (output_bb): Dump stmts streamed.
16727         * lto-section-out.c: Include print-tree.h
16728         (lto_begin_section): Dump sections created.
16729         (lto_output_decl_index): Dump decl encoded.
16730         * lto-streamer-out.c: Include print-tree.h
16731         (create_output_block): Dump output block created.
16732         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16733         (output_function): Dump function output.
16734         (output_constructor): Dump constructor streamed.
16735         (write_global_stream): Output indexes encoded.
16736         (produce_asm_for_decls): Dump streams encoded.
16737         * lto-streamer.c (streamer_dump_file): New global var.
16738         * lto-streamer.h (streamer_dump_file): Declare.
16739         * passes.c (ipa_write_summaries): Initialize streamer dump.
16740         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16741         in.
16743 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
16745         PR target/86048
16746         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16747         offsets for register save directives.  Emit a second batch of save
16748         directives, if need be, when the function accesses prior frames.
16750 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16752         * config/arc/fpu.md (fmasf4): Force operand to register.
16753         (fnmasf4): Likewise.
16755 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16757         * config/arc/arc-protos.h (arc_pad_return): Remove.
16758         * config/arc/arc.c (machine_function): Remove force_short_suffix
16759         and size_reason.
16760         (arc_print_operand): Adjust printing of '&'.
16761         (arc_verify_short): Remove conditional printing of short suffix.
16762         (arc_final_prescan_insn): Remove reference to size_reason.
16763         (pad_return): New function.
16764         (arc_reorg): Call pad_return.
16765         (arc_pad_return): Remove.
16766         (arc_init_machine_status): Remove reference to force_short_suffix.
16767         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16768         (attr length): When attribute iscompact is true force to 2
16769         regardless; in the case of maybe check if we want to force the
16770         instruction to have 4 bytes length.
16771         (nopv): Change it to generate 4 byte long nop as well.
16772         (blockage): New pattern.
16773         (simple_return): Remove call to arc_pad_return.
16774         (p_return_i): Likewise.
16776 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16778         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16780 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16782         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16783         ARC cores.
16785 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16787         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16788         for ARC700 and ARCv2.
16790 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
16792         PR target/86076
16793         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16794         operands[2] instead of operands[1].
16797 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16799         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16800         case, check whether the outer register overlaps an unallocatable
16801         register, not just whether it fits the required class.
16803 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16805         * poly-int.h (can_div_trunc_p): Add new overload in which all values
16806         are poly_ints.
16807         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16808         (memrefs_conflict_p): Likewise.
16809         (init_alias_analysis): Likewise.
16810         * cfgexpand.c (expand_debug_expr): Likewise.
16811         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16812         * cse.c (fold_rtx): Likewise.
16813         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16814         * expr.c (emit_block_move_hints): Likewise.
16815         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16816         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16817         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16818         (emit_group_store): Likewise.
16819         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16820         to read the PRE/POST_MODIFY increment.
16821         * calls.c (store_one_arg): Use strip_offset.
16822         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16823         poly_int_rtx_p.
16824         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16825         by a VEC_SELECT.
16826         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16827         (simplify_binary_operation_1): Extend CONST_INT handling to
16828         poly_int_rtx_p.
16829         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16830         than a HOST_WIDE_INT.
16831         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16832         poly_int64.
16833         (adjust_mems, add_stores): Update accodingly.
16834         (vt_canonicalize_addr): Track polynomial offsets.
16835         (emit_note_insn_var_location): Likewise.
16836         (vt_add_function_parameter): Likewise.
16837         (vt_initialize): Likewise.
16839 2018-06-12  Jeff Law  <law@redhat.com>
16841         * config.gcc (alpha*-*-freebsd*): Remove.
16842         * config/alpha/freebsd.h: Remove.
16844 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16846         PR other/69968
16847         * spellcheck-tree.c (levenshtein_distance): Rename to...
16848         (get_edit_distance): ...this, and update for underlying renaming.
16849         * spellcheck-tree.h (levenshtein_distance): Rename to...
16850         (get_edit_distance): ...this.
16851         * spellcheck.c (levenshtein_distance): Rename to...
16852         (get_edit_distance): ...this.  Convert from Levenshtein distance
16853         to Damerau-Levenshtein distance by supporting transpositions of
16854         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16855         "v_one_ago".
16856         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16857         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16858         update for underlying renaming.
16859         (selftest::levenshtein_distance_unit_test): Rename to...
16860         (selftest::test_get_edit_distance_unit): ...this, and update for
16861         underlying renaming.
16862         (selftest::test_find_closest_string): Add example from PR 69968
16863         where transposition helps
16864         (selftest::test_metric_conditions): Update for renaming.
16865         (selftest::test_metric_conditions): Likewise.
16866         (selftest::spellcheck_c_tests): Likewise.
16867         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16868         (get_edit_distance): ...this.
16869         (best_match::consider): Update for renaming.
16871 2018-06-12  Martin Sebor  <msebor@redhat.com>
16873         PR tree-optimization/85259
16874         * builtins.c (compute_objsize): Handle constant offsets.
16875         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16876         true iff a warning has been issued.
16877         * gimple.h (gimple_nonartificial_location): New function.
16878         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16879         gimple_nonartificial_location and handle -Wno-system-headers.
16880         (handle_builtin_stxncpy): Same.
16882 2018-06-12  Martin Sebor  <msebor@redhat.com>
16884         PR c/85931
16885         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16887 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16889         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16890         BUILTIN_VEC_XST entries for pointer to double and long long.
16892 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16894         PR target/85990
16895         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16896         Update comments.
16897         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16898         Likewise.
16900 2018-06-12  Martin Liska  <mliska@suse.cz>
16902         * doc/options.texi: Document IntegerRange.
16904 2018-06-12  Martin Liska  <mliska@suse.cz>
16906         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16907         * opt-functions.awk: Handle Deprecated flag.
16908         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16909         and report error.
16910         (read_cmdline_option): Report warning for a deprecated option.
16911         * opts.h (struct cl_option): Add new field cl_deprecated.
16912         (CL_ERR_DEPRECATED): New.
16914 2018-06-12  Martin Liska  <mliska@suse.cz>
16916         * doc/options.texi: Document Deprecated option flag.
16918 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16920         * config/arc/arc-arch.h (arc_extras): New enum.
16921         (arc_cpu_t):Add field extra.
16922         (arc_cpu_types): Consider the extras.
16923         * config/arc/arc-cpus.def: Add extras info.
16924         * config/arc/arc-opts.h (processor_type): Consider extra field.
16925         * config/arc/arc.c (arc_override_options): Handle extra field.
16927 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16929         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16930         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16931         field.
16932         * config/arc/arc.c (arc_init): Update pic warning.
16933         (irq_range): Update irq range parsing warnings.
16934         (arc_override_options): Update various warning messages.
16935         (arc_handle_aux_attribute): Likewise.
16937 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16939         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16941 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16943         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16944         compiler messages directives.
16946 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16948         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16949         * config/mips/mips-tables.opt: Regenerate.
16950         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16951         mips64r6.
16952         * doc/invoke.texi: Document -march=i6500.
16954 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16956         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16957         (i6400_gpmul): Add cpu_unit.
16958         (i6400_gpdiv): Likewise.
16959         (i6400_msa_add_d): Update reservations.
16960         (i6400_msa_int_add) Likewise.
16961         (i6400_msa_short_logic3) Likewise.
16962         (i6400_msa_short_logic2) Likewise.
16963         (i6400_msa_short_logic) Likewise.
16964         (i6400_msa_move) Likewise.
16965         (i6400_msa_cmp) Likewise.
16966         (i6400_msa_short_float2) Likewise.
16967         (i6400_msa_div_d) Likewise.
16968         (i6400_msa_long_logic1) Likewise.
16969         (i6400_msa_long_logic2) Likewise.
16970         (i6400_msa_mult) Likewise.
16971         (i6400_msa_long_float2) Likewise.
16972         (i6400_msa_long_float4) Likewise.
16973         (i6400_msa_long_float5) Likewise.
16974         (i6400_msa_long_float8) Likewise.
16975         (i6400_fpu_fadd): Include frint type.
16976         (i6400_fpu_store): New define_insn_reservation.
16977         (i6400_fpu_load): Likewise.
16978         (i6400_fpu_move): Likewise.
16979         (i6400_fpu_fcmp): Likewise.
16980         (i6400_fpu_fmadd): Likewise.
16981         (i6400_int_mult): Include imul3nc type and update reservation.
16982         (i6400_int_div): Include idiv3 type and update reservation.
16983         (i6400_int_load): Update to check type not move_type.
16984         (i6400_int_store): Likewise.
16985         (i6400_int_prefetch): Set zero latency.
16987 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16989         * gcc.c: Document new %@{...} sequence.
16990         (LINK_COMMAND_SPEC): Use it for the -L switches.
16991         (cpp_unique_options): Use it for the -I switches.
16992         (at_file_argbuf): New global variable.
16993         (in_at_file): Likewise.
16994         (alloc_args): Create at_file_argbuf.
16995         (clear_args): Truncate at_file_argbuf.
16996         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16997         (open_at_file): New function.
16998         (close_at_file): Likewise.
16999         (create_at_file): Delete.
17000         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
17001         <'o'>: Likewise.
17002         <'@'>: New case.
17003         (validate_switches_from_spec): Deal with %@{...} sequence.
17004         (validate_switches): Likewise.
17005         (driver::finalize): Call clear_args.
17007 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
17009         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
17011 2018-06-11  Martin Sebor  <msebor@redhat.com>
17013         * doc/invoke.texi (-Wall): List -Wc++17-compat.
17014         (Wno-class-memaccess): Add @opindex.
17015         (Wno-templates, Wno-multiple-inheritance): Same.
17016         (Wno-virtual-inheritance, Wno-namespaces): Same.
17017         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
17018         (Wno-format-overflow, Wno-format-truncation): Same.
17019         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
17020         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
17021         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
17022         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
17023         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
17024         (Wno-misspelled-isr): Same.
17026 2018-06-11  Martin Sebor  <msebor@redhat.com>
17028         * PR tree-optimization/86083
17029         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
17031 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
17033         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
17035 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
17037         PR target/85755
17038         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
17039         on the correct operand.
17040         (*movdi_internal64): Ditto.
17042 2018-06-11  Martin Liska  <mliska@suse.cz>
17044         PR tree-optimization/86089
17045         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
17047 2018-06-11  Julia Koval  <julia.koval@intel.com>
17049         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
17050         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
17051         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
17053 2018-06-11  Olivier Hainque  <hainque@adacore.com>
17055         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
17056         for Ada with strict dwarf2.
17058 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
17060         PR target/85755
17061         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
17062         addresses.
17064 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
17066         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
17068 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
17070         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
17071         TARGET_ELF.
17073 2018-06-08  Martin Liska  <mliska@suse.cz>
17075         * tree-cfg.h (debug_function): Fix argument type to match
17076         implementation.
17078 2018-06-08  Martin Liska  <mliska@suse.cz>
17080         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
17081         Remove usage of MPX-related (and removed) fields.
17082         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
17084 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
17086         * cfg.c (debug): Use TDF_NONE rather than 0.
17087         * cfghooks.c (debug): Likewise.
17088         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
17089         (struct dump_option_value_info): Convert to...
17090         (struct kv_pair): ...this template type.
17091         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
17092         rather than 0.
17093         (optinfo_verbosity_options): Likewise.
17094         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
17095         OPTGROUP_NONE.
17096         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
17097         than int for "optgroup_flags" param.
17098         (dump_generic_expr_loc): Use dump_flags_t rather than int for
17099         "dump_kind" param.
17100         (dump_dec): Likewise.
17101         (dump_finish): Use TDF_NONE rather than 0.
17102         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
17103         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
17104         than 0.  Update for change to option_ptr.
17105         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
17106         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
17107         0.  Update for changes to optinfo_verbosity_options and
17108         optgroup_options.
17109         (opt_info_switch_p): Convert optgroup_flags from int to
17110         optgroup_flags_t.
17111         (dump_basic_block): Use dump_flags_t rather than int
17112         for "dump_kind" param.
17113         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
17114         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
17115         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
17116         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
17117         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
17118         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
17119         TDF_NONE): Convert from macros to...
17120         (enum dump_flag): ...this new enum.
17121         (dump_flags_t): Update to use enum.
17122         (operator|, operator&, operator~, operator|=, operator&=):
17123         Implement for dump_flags_t.
17124         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
17125         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
17126         Convert from macros to...
17127         (enum optgroup_flag): ...this new enum.
17128         (optgroup_flags_t): New typedef.
17129         (operator|, operator|=): Implement for optgroup_flags_t.
17130         (struct dump_file_info): Convert field "alt_flags" to
17131         dump_flags_t.  Convert field "optgroup_flags" to
17132         optgroup_flags_t.
17133         (dump_basic_block): Use dump_flags_t rather than int for param.
17134         (dump_generic_expr_loc): Likewise.
17135         (dump_dec): Likewise.
17136         (dump_register): Convert param "optgroup_flags" to
17137         optgroup_flags_t.
17138         (opt_info_enable_passes): Likewise.
17139         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
17140         than 0.
17141         * gimple-pretty-print.c (debug): Likewise.
17142         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
17143         (merged_store_group::apply_stores): Likewise.
17144         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
17145         * gimple.c (verify_gimple_pp): Likewise.
17146         * graphite-poly.c (print_pbb_body): Likewise.
17147         * passes.c (pass_manager::register_one_dump_file): Convert
17148         local "optgroup_flags" to optgroup_flags_t.
17149         * print-tree.c (print_node): Use TDF_NONE rather than 0.
17150         (debug): Likewise.
17151         (debug_body): Likewise.
17152         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
17153         to optgroup_flags_t.
17154         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
17155         than 0.
17156         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
17157         (convert_mult_to_fma): Likewise.
17158         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
17159         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
17160         * tree-vect-data-refs.c (dump_lower_bound): Convert param
17161         "dump_kind" to dump_flags_t.
17163 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
17165         * config/rs6000/rs6000.c (min, max): Delete.
17167 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
17169         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
17170         -mabi=spe and -mabi=no-spe.
17172 2018-06-08  Martin Liska  <mliska@suse.cz>
17174         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
17175         where we expect an existing summary.
17177 2018-06-08  Martin Liska  <mliska@suse.cz>
17179         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
17180         * ipa-inline.h (estimate_edge_growth): Likewise.
17182 2018-06-08  Martin Liska  <mliska@suse.cz>
17184         * cgraph.c (function_version_hasher::hash): Use
17185         cgraph_node::get_uid ().
17186         (function_version_hasher::equal):
17187         * cgraph.h (cgraph_node::get_uid): New method.
17188         * ipa-inline.c (update_caller_keys): Use
17189         cgraph_node::get_uid ().
17190         (update_callee_keys): Likewise.
17191         * ipa-utils.c (searchc): Likewise.
17192         (ipa_reduced_postorder): Likewise.
17193         * lto-cgraph.c (input_node): Likewise.
17194         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
17195         * symbol-summary.h (symtab_insertion): Likewise.
17196         (symtab_removal): Likewise.
17197         (symtab_duplication): Likewise.
17198         * tree-pretty-print.c (dump_function_header): Likewise.
17199         * tree-sra.c (convert_callers_for_node): Likewise.
17201 2018-06-08  Martin Liska  <mliska@suse.cz>
17203         * cgraph.c (symbol_table::create_edge): Always assign a new
17204         unique number.
17205         (symbol_table::free_edge): Do not recycle numbers.
17206         * cgraph.h (cgraph_edge::get): New method.
17207         * symbol-summary.h (symtab_removal): Use it.
17208         (symtab_duplication): Likewise.
17209         (call_summary::hashable_uid): Remove.
17211 2018-06-08  Martin Liska  <mliska@suse.cz>
17213         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
17214         (initialize_growth_caches): Remove.
17215         (free_growth_caches): Likewise.
17216         (do_estimate_edge_time): Use edge_growth_cache.
17217         (do_estimate_edge_size): Likewise.
17218         (do_estimate_edge_hints): Likewise.
17219         * ipa-inline.c (reset_edge_caches): Likewise.
17220         (recursive_inlining): Likewise.
17221         (inline_small_functions): Likewise.
17222         * ipa-inline.h (initialize_growth_caches): Remove.
17223         (estimate_edge_size): Likewise.
17224         (estimate_edge_time): Likewise.
17225         (estimate_edge_hints): Likewise.
17226         (reset_edge_growth_cache): Likewise.
17227         * symbol-summary.h (call_summary::remove): New method.
17229 2018-06-08  Martin Liska  <mliska@suse.cz>
17231         * ipa-cp.c (class edge_clone_summary): New summary.
17232         (grow_edge_clone_vectors): Remove.
17233         (ipcp_edge_duplication_hook): Remove.
17234         (class edge_clone_summary_t): New call_summary class.
17235         (ipcp_edge_removal_hook): Remove.
17236         (edge_clone_summary_t::duplicate): New function.
17237         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
17238         (create_specialized_node): Likewise.
17239         (ipcp_driver): Initialize edge_clone_summaries and do not
17240         register hooks.
17242 2018-06-08  Martin Liska  <mliska@suse.cz>
17244         * symbol-summary.h (get): New function.
17245         (call_summary::m_initialize_when_cloning): New class member.
17247 2018-06-08  Martin Liska  <mliska@suse.cz>
17249         * cgraph.c (cgraph_node::remove): Do not recycle uid.
17250         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
17251         (symbol_table::allocate_cgraph_symbol): Do not set uid.
17252         * passes.c (uid_hash_t): Record removed_nodes by their uids.
17253         (remove_cgraph_node_from_order): Use the removed_nodes set.
17254         (do_per_function_toporder): Likwise.
17255         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
17256         instead of summary_uid.
17257         (symtab_removal): Likewise.
17258         (symtab_duplication): Likewise.
17260 2018-06-08  Martin Liska  <mliska@suse.cz>
17262         * ipa-cp.c (ipcp_store_bits_results): Use
17263         ipcp_transformation_sum.
17264         (ipcp_store_vr_results): Likewise.
17265         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
17266         to ...
17267         (ipcp_transformation_initialize): ... this.
17268         (ipa_set_node_agg_value_chain):
17269         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
17270         (write_ipcp_transformation_info): Likewise.
17271         (read_ipcp_transformation_info): Likewise.
17272         (ipcp_update_bits): Likewise.
17273         (ipcp_update_vr): Likewise.
17274         (ipcp_transform_function): Likewise.
17275         * ipa-prop.h: Rename ipcp_transformation_summary to
17276         ipcp_transformation.
17277         (class ipcp_transformation_t): New function summary.
17278         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
17279         (ipa_get_agg_replacements_for_node): Likewise.
17281 2018-06-08  Martin Liska  <mliska@suse.cz>
17283         * ipa-pure-const.c (struct funct_state_d): Do it class instead
17284         of struct.
17285         (class funct_state_summary_t): New function_summary class.
17286         (has_function_state): Remove.
17287         (get_function_state): Likewise.
17288         (set_function_state): Likewise.
17289         (add_new_function): Likewise.
17290         (funct_state_summary_t::insert): New function.
17291         (duplicate_node_data): Remove.
17292         (remove_node_data): Remove.
17293         (funct_state_summary_t::duplicate): New function.
17294         (register_hooks): Create new funct_state_summaries.
17295         (pure_const_generate_summary): Use it.
17296         (pure_const_write_summary): Likewise.
17297         (pure_const_read_summary): Likewise.
17298         (propagate_pure_const): Likewise.
17299         (propagate_nothrow): Likewise.
17300         (dump_malloc_lattice): Likewise.
17301         (propagate_malloc): Likewise.
17302         (execute): Do not register hooks, just remove summary
17303         instead.
17304         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
17305         constructor.
17307 2018-06-08  Martin Liska  <mliska@suse.cz>
17309         * ipa-reference.c (remove_node_data): Remove.
17310         (duplicate_node_data): Likewise.
17311         (class ipa_ref_var_info_summary_t): New class.
17312         (class ipa_ref_opt_summary_t): Likewise.
17313         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
17314         (get_reference_optimization_summary): Use
17315         ipa_ref_opt_sum_summaries.
17316         (set_reference_vars_info): Remove.
17317         (set_reference_optimization_summary): Likewise.
17318         (ipa_init): Create summaries.
17319         (init_function_info): Use function summary.
17320         (ipa_ref_opt_summary_t::duplicate): New function.
17321         (ipa_ref_opt_summary_t::remove): New function.
17322         (get_read_write_all_from_node): Fix GNU coding style.
17323         (propagate): Use function summary.
17324         (write_node_summary_p): Fix GNU coding style.
17325         (stream_out_bitmap): Likewise.
17326         (ipa_reference_read_optimization_summary): Use function summary.
17327         (ipa_reference_c_finalize): Do not release hooks.
17329 2018-06-08  Martin Liska  <mliska@suse.cz>
17331         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
17332         (analyze_function_body): Extract multiple calls of get_create.
17333         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
17334         * ipa-inline.c (recursive_inlining): Use ::get method.
17335         * ipa-inline.h (estimate_edge_growth): Likewise.
17337 2018-06-08  Martin Liska  <mliska@suse.cz>
17339         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
17340         HSA_INVALID.
17341         (hsa_function_summary::hsa_function_summary): Use the new enum
17342         value.
17343         (hsa_gpu_implementation_p): Use hsa_summaries::get.
17344         * hsa-gen.c (hsa_get_host_function): Likewise.
17345         (get_brig_function_name): Likewise.
17346         * ipa-hsa.c (process_hsa_functions): Likewise.
17347         (ipa_hsa_write_summary): Likewise.
17348         * symbol-summary.h (symtab_duplication): Use ::get function/
17349         (get): New function.
17351 2018-06-08  Martin Liska  <mliska@suse.cz>
17353         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
17354         of get.
17355         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
17356         (hsa_register_kernel): Likewise.
17357         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
17358         * hsa-gen.c (hsa_get_host_function): Likewise.
17359         (get_brig_function_name): Likewise.
17360         (generate_hsa): Likewise.
17361         (pass_gen_hsail::execute): Likewise.
17362         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
17363         (devirtualization_time_bonus): Likewise.
17364         (ipcp_propagate_stage): Likewise.
17365         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
17366         (edge_set_predicate): Likewise.
17367         (evaluate_conditions_for_known_args): Likewise.
17368         (evaluate_properties_for_edge): Likewise.
17369         (ipa_fn_summary::reset): Likewise.
17370         (ipa_fn_summary_t::duplicate): Likewise.
17371         (dump_ipa_call_summary): Likewise.
17372         (ipa_dump_fn_summary): Likewise.
17373         (analyze_function_body): Likewise.
17374         (compute_fn_summary): Likewise.
17375         (estimate_edge_devirt_benefit): Likewise.
17376         (estimate_edge_size_and_time): Likewise.
17377         (estimate_calls_size_and_time): Likewise.
17378         (estimate_node_size_and_time): Likewise.
17379         (inline_update_callee_summaries): Likewise.
17380         (remap_edge_change_prob): Likewise.
17381         (remap_edge_summaries): Likewise.
17382         (ipa_merge_fn_summary_after_inlining): Likewise.
17383         (ipa_update_overall_fn_summary): Likewise.
17384         (read_ipa_call_summary): Likewise.
17385         (inline_read_section): Likewise.
17386         (write_ipa_call_summary): Likewise.
17387         (ipa_fn_summary_write): Likewise.
17388         (ipa_free_fn_summary): Likewise.
17389         * ipa-hsa.c (process_hsa_functions): Likewise.
17390         (ipa_hsa_write_summary): Likewise.
17391         (ipa_hsa_read_section): Likewise.
17392         * ipa-icf.c (sem_function::merge): Likewise.
17393         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
17394         (do_estimate_edge_time): Likewise.
17395         (estimate_size_after_inlining): Likewise.
17396         (estimate_growth): Likewise.
17397         (growth_likely_positive): Likewise.
17398         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
17399         (inline_call): Likewise.
17400         * ipa-inline.c (caller_growth_limits): Likewise.
17401         (can_inline_edge_p): Likewise.
17402         (can_inline_edge_by_limits_p): Likewise.
17403         (compute_uninlined_call_time): Likewise.
17404         (compute_inlined_call_time): Likewise.
17405         (want_inline_small_function_p): Likewise.
17406         (edge_badness): Likewise.
17407         (update_caller_keys): Likewise.
17408         (update_callee_keys): Likewise.
17409         (recursive_inlining): Likewise.
17410         (inline_small_functions): Likewise.
17411         (inline_to_all_callers_1): Likewise.
17412         (dump_overall_stats): Likewise.
17413         (early_inline_small_functions): Likewise.
17414         (early_inliner): Likewise.
17415         * ipa-inline.h (estimate_edge_growth): Likewise.
17416         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
17417         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
17418         * ipa-prop.h (IPA_NODE_REF): Likewise.
17419         (IPA_EDGE_REF): Likewise.
17420         * ipa-pure-const.c (malloc_candidate_p): Likewise.
17421         (propagate_malloc): Likewise.
17422         * ipa-split.c (execute_split_functions): Likewise.
17423         * symbol-summary.h: Rename get to get_create.
17424         (get): Likewise.
17425         (get_create): Likewise.
17426         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
17428 2018-06-08  Martin Liska  <mliska@suse.cz>
17430         * symbol-summary.h (release): Move definition out of class
17431         declaration.
17432         (symtab_removal): Likewise.
17433         (symtab_duplication): Likewise.
17435 2018-06-08  Martin Liska  <mliska@suse.cz>
17437         * symbol-summary.h (function_summary): Move constructor
17438         implementation out of class declaration.
17439         (release): Likewise.
17440         (symtab_insertion): Likewise.
17441         (symtab_removal): Likewise.
17442         (symtab_duplication): Likewise.
17443         (get): Likewise.
17445 2018-06-08  Martin Liska  <mliska@suse.cz>
17447         * Makefile.in: Remove support for MPX (macros, related functions,
17448         fields in cgraph_node, ...).
17449         * builtin-types.def (BT_BND): Likewise.
17450         (BT_FN_BND_CONST_PTR): Likewise.
17451         (BT_FN_CONST_PTR_BND): Likewise.
17452         (BT_FN_VOID_PTR_BND): Likewise.
17453         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
17454         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
17455         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
17456         (expand_builtin_mempcpy_with_bounds): Likewise.
17457         (expand_builtin_memset_with_bounds): Likewise.
17458         (expand_builtin_memset_args): Likewise.
17459         (std_expand_builtin_va_start): Likewise.
17460         (expand_builtin): Likewise.
17461         (expand_builtin_with_bounds): Likewise.
17462         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
17463         (DEF_LIB_BUILTIN_CHKP): Likewise.
17464         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
17465         (DEF_CHKP_BUILTIN): Likewise.
17466         (BUILT_IN_MEMCPY): Likewise.
17467         (BUILT_IN_MEMMOVE): Likewise.
17468         (BUILT_IN_MEMPCPY): Likewise.
17469         (BUILT_IN_MEMSET): Likewise.
17470         (BUILT_IN_STPCPY): Likewise.
17471         (BUILT_IN_STRCAT): Likewise.
17472         (BUILT_IN_STRCHR): Likewise.
17473         (BUILT_IN_STRCPY): Likewise.
17474         (BUILT_IN_STRLEN): Likewise.
17475         (BUILT_IN_MEMCPY_CHK): Likewise.
17476         (BUILT_IN_MEMMOVE_CHK): Likewise.
17477         (BUILT_IN_MEMPCPY_CHK): Likewise.
17478         (BUILT_IN_MEMSET_CHK): Likewise.
17479         (BUILT_IN_STPCPY_CHK): Likewise.
17480         (BUILT_IN_STRCAT_CHK): Likewise.
17481         (BUILT_IN_STRCPY_CHK): Likewise.
17482         * calls.c (store_bounds): Likewise.
17483         (emit_call_1): Likewise.
17484         (special_function_p): Likewise.
17485         (maybe_warn_nonstring_arg): Likewise.
17486         (initialize_argument_information): Likewise.
17487         (finalize_must_preallocate): Likewise.
17488         (compute_argument_addresses): Likewise.
17489         (expand_call): Likewise.
17490         * cfgexpand.c (expand_call_stmt): Likewise.
17491         (expand_return): Likewise.
17492         (expand_gimple_stmt_1): Likewise.
17493         (pass_expand::execute): Likewise.
17494         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
17495         (cgraph_node::remove): Likewise.
17496         (cgraph_node::dump): Likewise.
17497         (cgraph_node::verify_node): Likewise.
17498         * cgraph.h (chkp_function_instrumented_p): Likewise.
17499         (symtab_node::get_alias_target): Likewise.
17500         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
17501         (cgraph_local_p): Likewise.
17502         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
17503         (cgraph_edge::rebuild_references): Likewise.
17504         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
17505         (walk_polymorphic_call_targets): Likewise.
17506         (cgraph_node::expand_thunk): Likewise.
17507         (symbol_table::output_weakrefs): Likewise.
17508         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
17509         (ix86_handle_option): Likewise.
17510         * config/i386/constraints.md: Likewise.
17511         * config/i386/i386-builtin-types.def (BND): Likewise.
17512         (VOID): Likewise.
17513         (PVOID): Likewise.
17514         (ULONG): Likewise.
17515         * config/i386/i386-builtin.def (BDESC_END): Likewise.
17516         (BDESC_FIRST): Likewise.
17517         (BDESC): Likewise.
17518         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
17519         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
17520         * config/i386/i386.c (enum reg_class): Likewise.
17521         (ix86_target_string): Likewise.
17522         (ix86_option_override_internal): Likewise.
17523         (ix86_conditional_register_usage): Likewise.
17524         (ix86_valid_target_attribute_inner_p): Likewise.
17525         (ix86_set_indirect_branch_type): Likewise.
17526         (ix86_set_current_function): Likewise.
17527         (ix86_function_arg_regno_p): Likewise.
17528         (init_cumulative_args): Likewise.
17529         (ix86_function_arg_advance): Likewise.
17530         (ix86_function_arg): Likewise.
17531         (ix86_pass_by_reference): Likewise.
17532         (ix86_function_value_regno_p): Likewise.
17533         (ix86_function_value_1): Likewise.
17534         (ix86_function_value_bounds): Likewise.
17535         (ix86_return_in_memory): Likewise.
17536         (ix86_setup_incoming_vararg_bounds): Likewise.
17537         (ix86_va_start): Likewise.
17538         (indirect_thunk_need_prefix): Likewise.
17539         (print_reg): Likewise.
17540         (ix86_print_operand): Likewise.
17541         (ix86_expand_call): Likewise.
17542         (ix86_output_function_return): Likewise.
17543         (reg_encoded_number): Likewise.
17544         (BDESC_VERIFYS): Likewise.
17545         (ix86_init_mpx_builtins): Likewise.
17546         (ix86_init_builtins): Likewise.
17547         (ix86_emit_cmove): Likewise.
17548         (ix86_emit_move_max): Likewise.
17549         (ix86_expand_builtin): Likewise.
17550         (ix86_builtin_mpx_function): Likewise.
17551         (ix86_get_arg_address_for_bt): Likewise.
17552         (ix86_load_bounds): Likewise.
17553         (ix86_store_bounds): Likewise.
17554         (ix86_load_returned_bounds): Likewise.
17555         (ix86_store_returned_bounds): Likewise.
17556         (ix86_class_likely_spilled_p): Likewise.
17557         (ix86_hard_regno_mode_ok): Likewise.
17558         (x86_order_regs_for_local_alloc): Likewise.
17559         (ix86_mitigate_rop): Likewise.
17560         (ix86_bnd_prefixed_insn_p): Likewise.
17561         (ix86_mpx_bound_mode): Likewise.
17562         (ix86_make_bounds_constant): Likewise.
17563         (ix86_initialize_bounds): Likewise.
17564         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
17565         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
17566         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
17567         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
17568         (TARGET_CHKP_BOUND_MODE): Likewise.
17569         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
17570         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
17571         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
17572         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
17573         * config/i386/i386.h (TARGET_MPX): Likewise.
17574         (TARGET_MPX_P): Likewise.
17575         (VALID_BND_REG_MODE): Likewise.
17576         (FIRST_BND_REG): Likewise.
17577         (LAST_BND_REG): Likewise.
17578         (enum reg_class): Likewise.
17579         (BND_REG_P): Likewise.
17580         (BND_REGNO_P): Likewise.
17581         (BNDmode): Likewise.
17582         (ADJUST_INSN_LENGTH): Likewise.
17583         * config/i386/i386.md: Likewise.
17584         * config/i386/i386.opt: Likewise.
17585         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
17586         (defined): Likewise.
17587         (LINK_MPX): Likewise.
17588         (MPX_SPEC): Likewise.
17589         (LIBMPX_SPEC): Likewise.
17590         (LIBMPXWRAPPERS_SPEC): Likewise.
17591         (CHKP_SPEC): Likewise.
17592         * config/i386/predicates.md: Likewise.
17593         * dbxout.c (dbxout_type): Likewise.
17594         * doc/extend.texi: Likewise.
17595         * doc/invoke.texi: Likewise.
17596         * doc/md.texi: Likewise.
17597         * doc/tm.texi: Likewise.
17598         * doc/tm.texi.in: Likewise.
17599         * dwarf2out.c (is_base_type): Likewise.
17600         (gen_formal_types_die): Likewise.
17601         (gen_subprogram_die): Likewise.
17602         (gen_type_die_with_usage): Likewise.
17603         (gen_decl_die): Likewise.
17604         (dwarf2out_late_global_decl): Likewise.
17605         * expr.c (expand_assignment): Likewise.
17606         (emit_storent_insn): Likewise.
17607         (store_expr_with_bounds): Likewise.
17608         (store_expr): Likewise.
17609         (expand_expr_real_1): Likewise.
17610         * expr.h (store_expr_with_bounds): Likewise.
17611         * function.c (use_register_for_decl): Likewise.
17612         (struct bounds_parm_data): Likewise.
17613         (assign_parms_augmented_arg_list): Likewise.
17614         (assign_parm_find_entry_rtl): Likewise.
17615         (assign_parm_is_stack_parm): Likewise.
17616         (assign_parm_load_bounds): Likewise.
17617         (assign_bounds): Likewise.
17618         (assign_parms): Likewise.
17619         (expand_function_start): Likewise.
17620         * gcc.c (CHKP_SPEC): Likewise.
17621         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
17622         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
17623         (wrestrict_dom_walker::check_call): Likewise.
17624         * gimple.c (gimple_build_call_from_tree): Likewise.
17625         * gimple.h (enum gf_mask): Likewise.
17626         (gimple_call_with_bounds_p): Likewise.
17627         (gimple_call_set_with_bounds): Likewise.
17628         * gimplify.c (gimplify_init_constructor): Likewise.
17629         * ipa-cp.c (initialize_node_lattices): Likewise.
17630         (propagate_constants_across_call): Likewise.
17631         (find_more_scalar_values_for_callers_subset): Likewise.
17632         * ipa-hsa.c (process_hsa_functions): Likewise.
17633         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
17634         * ipa-icf.c (sem_function::merge): Likewise.
17635         * ipa-inline.c (early_inliner): Likewise.
17636         * ipa-pure-const.c (warn_function_noreturn): Likewise.
17637         (warn_function_cold): Likewise.
17638         (propagate_pure_const): Likewise.
17639         * ipa-ref.h (enum GTY): Likewise.
17640         * ipa-split.c (find_retbnd): Likewise.
17641         (consider_split): Likewise.
17642         (split_function): Likewise.
17643         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
17644         * ipa.c (walk_polymorphic_call_targets): Likewise.
17645         (symbol_table::remove_unreachable_nodes): Likewise.
17646         (process_references): Likewise.
17647         (cgraph_build_static_cdtor_1): Likewise.
17648         * lto-cgraph.c (lto_output_node): Likewise.
17649         (output_refs): Likewise.
17650         (compute_ltrans_boundary): Likewise.
17651         (input_overwrite_node): Likewise.
17652         (input_node): Likewise.
17653         (input_cgraph_1): Likewise.
17654         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
17655         * passes.c (pass_manager::execute_early_local_passes): Likewise.
17656         (class pass_chkp_instrumentation_passes): Likewise.
17657         (make_pass_chkp_instrumentation_passes): Likewise.
17658         * passes.def: Likewise.
17659         * rtl.h (struct GTY): Likewise.
17660         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
17661         * stor-layout.c (layout_type): Likewise.
17662         * symtab.c: Likewise.
17663         * target.def: Likewise.
17664         * targhooks.c (default_chkp_bound_type): Likewise.
17665         (default_chkp_bound_mode): Likewise.
17666         (default_builtin_chkp_function): Likewise.
17667         (default_chkp_function_value_bounds): Likewise.
17668         (default_chkp_make_bounds_constant): Likewise.
17669         (default_chkp_initialize_bounds): Likewise.
17670         * targhooks.h (default_chkp_bound_type): Likewise.
17671         (default_chkp_bound_mode): Likewise.
17672         (default_builtin_chkp_function): Likewise.
17673         (default_chkp_function_value_bounds): Likewise.
17674         (default_chkp_make_bounds_constant): Likewise.
17675         (default_chkp_initialize_bounds): Likewise.
17676         * toplev.c (compile_file): Likewise.
17677         (process_options): Likewise.
17678         * tree-core.h (DEF_BUILTIN): Likewise.
17679         (DEF_BUILTIN_CHKP): Likewise.
17680         * tree-inline.c (declare_return_variable): Likewise.
17681         (remap_gimple_stmt): Likewise.
17682         (copy_bb): Likewise.
17683         (initialize_inlined_parameters): Likewise.
17684         (expand_call_inline): Likewise.
17685         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17686         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17687         (make_pass_ipa_chkp_produce_thunks): Likewise.
17688         (make_pass_chkp): Likewise.
17689         (make_pass_chkp_opt): Likewise.
17690         (make_pass_chkp_instrumentation_passes): Likewise.
17691         * tree-pretty-print.c (dump_generic_node): Likewise.
17692         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17693         * tree-ssa-dce.c (propagate_necessity): Likewise.
17694         (eliminate_unnecessary_stmts): Likewise.
17695         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17696         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17697         * tree-ssa-sccvn.h: Likewise.
17698         * tree-ssa-strlen.c (get_string_length): Likewise.
17699         (valid_builtin_call): Likewise.
17700         (adjust_last_stmt): Likewise.
17701         (handle_builtin_strchr): Likewise.
17702         (handle_builtin_strcpy): Likewise.
17703         (handle_builtin_stxncpy): Likewise.
17704         (handle_builtin_memcpy): Likewise.
17705         (handle_builtin_strcat): Likewise.
17706         (strlen_check_and_optimize_stmt): Likewise.
17707         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17708         * tree-streamer-in.c: Likewise.
17709         * tree-streamer.c (record_common_node): Likewise.
17710         * tree.c (tree_code_size): Likewise.
17711         (wide_int_to_tree_1): Likewise.
17712         (type_contains_placeholder_1): Likewise.
17713         (build_common_tree_nodes): Likewise.
17714         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17715         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17716         (POINTER_BOUNDS_P): Likewise.
17717         (BOUNDED_TYPE_P): Likewise.
17718         (BOUNDED_P): Likewise.
17719         (CALL_WITH_BOUNDS_P): Likewise.
17720         (pointer_bounds_type_node): Likewise.
17721         * value-prof.c (gimple_ic): Likewise.
17722         * var-tracking.c (vt_add_function_parameters): Likewise.
17723         * varasm.c (make_decl_rtl): Likewise.
17724         (assemble_start_function): Likewise.
17725         (output_constant): Likewise.
17726         (maybe_assemble_visibility): Likewise.
17727         * varpool.c (ctor_for_folding): Likewise.
17728         * chkp-builtins.def: Remove.
17729         * ipa-chkp.c: Remove.
17730         * ipa-chkp.h: Remove.
17731         * rtl-chkp.c: Remove.
17732         * rtl-chkp.h: Remove.
17733         * tree-chkp-opt.c: Remove.
17734         * tree-chkp.c: Remove.
17735         * tree-chkp.h: Remove.
17737 2018-06-07  Carl Love  <cel@us.ibm.com>
17739         * config/rs6000/vsx.md (vextract_fp_from_shorth,
17740         vextract_fp_from_shortl): Add BE support.
17742 2018-06-07  Paul Koning  <ni1d@arrl.net>
17744         * compare-elim.c (try_merge_compare): Don't merge compare if
17745         address contains a side effect.
17746         (try_eliminate_compare): Likewise.
17748 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
17750         * config.gcc: Support "tremont".
17751         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17752         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17753         PROCESSOR_TREMONT.
17754         * config/i386/i386.c (m_TREMONT): Define.
17755         (processor_target_table): Add "tremont".
17756         (PTA_TREMONT): Define.
17757         (ix86_lea_outperforms): Add TARGET_TREMONT.
17758         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17759         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17760         and M_INTEL_GOLDMONT_PLUS.
17761         (fold_builtin_cpu): Add "tremont".
17762         (ix86_add_stmt_cost): Add TARGET_TREMONT.
17763         (ix86_option_override_internal): Add "tremont".
17764         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17765         (processor_type): Add PROCESSOR_TREMONT.
17766         * config/i386/x86-tune.def: Add m_TREMONT.
17767         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17769 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17771         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17772         symbol defined for msp430i* devices to be lower case.
17774 2018-06-07  Richard Biener  <rguenther@suse.de>
17776         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17777         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
17778         Properly wrap signed arithmetic if overflow wraps.
17780 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
17782         PR tree-optimization/69615
17783         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17784         of a cast from a same precision integral SSA_NAME in a bb dominated
17785         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
17786         cast to utype if rhs2 has already a compatible type.
17788 2018-06-07  Richard Biener  <rguenther@suse.de>
17790         PR tree-optimization/85935
17791         * graphite-scop-detection.c (find_params_in_bb): Analyze
17792         condition operands with respect to the correct loop.  Assert
17793         the analysis doesn't fail.
17795 2018-06-04  Carl Love  <cel@us.ibm.com>
17797         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17798         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
17799         as it is slightly cheaper.
17800         (first_match_or_eos_index_<mode>):
17801         Calculate index using natural element order.
17802         (first_match_index_<mode>):
17803         Calculate index using natural element order.
17804         (first_match_or_eos_index_<mode>):
17805         Calculate index using natural order.
17806         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17807         for BE and LE modes.
17808         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17809         P9V_BUILTIN_VCLZLSBB_V16QI.
17810         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17811         specific.
17813 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17815         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17816         indentation and line wrap for many prototypes.  Add missing
17817         @smallexample directives around block of prototypes for vec_xl and
17818         vec_xst.
17820 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17822         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17823         track if we pass or return IEEE 128-bit floating point.
17824         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17825         C++ mangling that is compatible with GCC 8.1.
17826         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17827         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17828         floating point types.
17829         (rs6000_function_arg_advance_1): Likewise.
17830         (rs6000_mangle_type): Optionally generate mangled names that match
17831         what GCC 8.1 generated for IEEE 128-bit floating point types.
17832         (rs6000_globalize_decl_name): If we have an external function that
17833         passes or returns IEEE 128-bit types, generate a weak reference
17834         from the mangled name used in GCC 8.1 to the current mangled
17835         name.
17836         (rs6000_init_builtins): Make __ibm128 use the long double type if
17837         long double is IBM extended double.  Make __float128 use the long
17838         double type if long double is IEEE 128-bit.
17840         PR target/85657
17841         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17842         macro for __ibm128 built-in functions.
17843         (PACK_IF): Add __ibm128 pack/unpack functions.
17844         (UNPACK_IF): Likewise.
17845         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17846         enable long double built-in functions if long double is IEEE
17847         128-bit floating point.
17848         (rs6000_invalid_builtin): Update long double built-in function
17849         error message.
17850         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17851         functions, adjust the built-in function to use the long double
17852         built-in function if __ibm128 and long double are the same type.
17853         * doc/extend.texi (PowerPC builtins): Update documention for
17854         __builtin_{,un}pack_longdouble.  Add documentation for
17855         __builtin_{,un}pack_ibm128.
17857 2018-06-06  Jim Wilson  <jimw@sifive.com>
17859         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17860         (struct machine_function): New field interrupt_mode.
17861         (riscv_handle_type_attribute): New function.  Add forward declaration.
17862         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17863         (riscv_expand_epilogue): Check interrupt_mode field.
17864         (riscv_set_current_function): Check interrupt attribute args and
17865         set interrupt_mode field.
17866         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17867         (riscv_sret, riscv_uret): New.
17868         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17869         new arguments to interrupt attribute.
17871 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17873         PR target/63177
17874         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17875         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17877 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17879         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17880         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17881         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17882         several redundant entries.
17884 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17886         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17887         type from "rtx" to "rtx_insn *".
17888         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17889         for local "call_insn", removing cast.
17890         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17892 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17894         PR tree-optimization/86066
17895         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17896         for BIT_INSERT_EXPR stores.
17898 2018-06-06  Richard Biener  <rguenther@suse.de>
17900         PR tree-optimization/86062
17901         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17902         component refs ontop
17903         of to be offsetted base.
17905 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17907         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17908         to be static and remove check on interrupt attribute name.
17910 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17912         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17913         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17915 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17917         PR target/79924
17918         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17919         second argument.
17920         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17921         Remove second argument, change how error is called.
17922         (aarch64_layout_arg): Remove second argument from
17923         aarch64_err_no_fpadvsimd call.
17924         (aarch64_init_cumulative_args): Ditto.
17925         (aarch64_gimplify_va_arg_expr): Ditto.
17926         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17928 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17930         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17931         (*simple_return_indirect_internal<mode>): Rename from
17932         simple_return_indirect_internal.  Use W mode iterator.
17933         (rstorssp): New expander.
17934         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17935         (clrssbsy): New expander.
17936         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17938 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17940         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17941         __typeof__.
17942         (cmse_check_pointed_object): Likewise.
17944 2018-06-05  Martin Liska  <mliska@suse.cz>
17946         PR gcov-profile/47618
17947         * doc/invoke.texi: Document how -fprofile-dir format
17948         is extended.
17950 2018-06-05  Richard Biener  <rguenther@suse.de>
17952         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17953         removal pretend DOM info isn't available so we do not update
17954         it and only remove edges, not dominated blocks.  Actually free
17955         DOM info in case we removed something.  Remove unreachable blocks.
17956         (mfb_keep_latches): Work with either DOM info or marked backedges.
17957         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17958         first.  Mark backedges if DOM info isn't available.
17959         (Re-)compute DOM info after cleanup_control_flow_pre.
17961 2018-06-05  Richard Biener  <rguenther@suse.de>
17963         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17964         (locus_discrim_hasher::hash): Adjust.
17965         (locus_discrim_hasher::equal): Likewise.
17966         (next_discriminator_for_locus): Work on line directly.
17967         (same_line_p): Pass in expanded locus1 as well.
17968         (assign_discriminators): Avoid redundant location expansions.
17970 2018-06-05  Richard Biener  <rguenther@suse.de>
17972         PR tree-optimization/86046
17973         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17974         if required after clearing TREE_ADDRESSABLE.
17976 2018-06-05  Richard Biener  <rguenther@suse.de>
17978         PR tree-optimization/86047
17979         * tree-ssa-loop.c (for_each_index): Glob handling of all
17980         decls and constants and really handle all of them.
17982 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17984         PR target/81497
17985         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17986         qualifier_void_pointer and qualifier_const_void_pointer.
17987         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17988         (arm_init_builtins): Handle the above.
17989         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17990         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17991         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17992         void intrinsics.
17994 2018-06-05  Martin Liska  <mliska@suse.cz>
17996         * auto-profile.c (read_autofdo_file): Do not use
17997         gcov_ctr_summary struct.
17998         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17999         * coverage.c (struct counts_entry): Likewise.
18000         (read_counts_file): Read just single summary entry.
18001         (get_coverage_counts): Use gcov_summary struct.
18002         * coverage.h (get_coverage_counts): Likewise.
18003         * gcov-dump.c (dump_working_sets): Likewise.
18004         (tag_summary): Dump just single summary.
18005         * gcov-io.c (gcov_write_summary): Write just histogram
18006         summary.
18007         (gcov_read_summary): Read just single summary.
18008         (compute_working_sets): Use gcov_summary struct.
18009         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
18010         of GCOV_COUNTERS_SUMMABLE.
18011         (GCOV_COUNTERS_SUMMABLE): Remove.
18012         (GCOV_FIRST_VALUE_COUNTER): Replace with
18013         GCOV_COUNTER_V_INTERVAL.
18014         (struct gcov_ctr_summary): Remove.
18015         (struct gcov_summary): Directly use fields of former
18016         gcov_ctr_summary.
18017         (compute_working_sets): Use gcov_summary struct.
18018         * gcov.c (read_count_file): Do not use ctrs fields.
18019         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
18020         struct.
18021         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
18022         struct.
18023         * profile.c: Likewise.
18024         * profile.h: Likewise.
18026 2018-06-05  Martin Liska  <mliska@suse.cz>
18028         PR gcov-profile/84846
18029         * gcov.c (output_lines): Print working directory only
18030         in intermediate format.
18032 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
18034         * config/s390/s390-builtin-types.def: Add void function type.
18035         * config/s390/s390-builtins.def: Use the function type for the
18036         tbeginc builtin.
18038 2018-06-04  Jim Wilson  <jimw@sifive.com>
18040         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
18041         to int.
18042         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
18043         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
18044         handle EH_RETURN_DATA_REGNO registers properly.
18045         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
18046         (riscv_expand_epilogue): Update comment.  Change argument name and
18047         type.  Update code to use new name and type.  Pass new args to
18048         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
18049         EXCEPTION_RETURN.
18050         * config/riscv/riscv.md (NORMAL_RETURN): New.
18051         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
18052         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
18053         (eh_return): Call gen_eh_return_internal and emit barrier.
18054         (eh_return_internal): Call riscv_expand_epilogue.
18056 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
18058         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
18059         bit_insertion field and declare can_be_merged_into method.
18060         (merged_store_group::can_be_merged_into): New method.
18061         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
18062         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
18063         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
18065 2018-06-04  Richard Biener  <rguenther@suse.de>
18067         PR tree-optimization/85955
18068         * builtins.c (fold_builtin_sincos): Convert pointers to
18069         destination to appropriate type before dereferencing.
18071 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
18073         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
18075 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
18077         * expr.c (expand_expr_real_1): Force the operand into memory if
18078         its TYPE_MODE is BLKmode and if there is no integer mode for
18079         the number of bits being extracted.
18081 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
18083         PR target/85832
18084         PR target/86036
18085         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18086         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
18088 2018-06-04  Richard Biener  <rguenther@suse.de>
18090         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
18091         (cleanup_tree_cfg_noloop): ... single caller.  Do
18092         start_recording_case_labels later.
18094 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
18096         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
18097         to _IMMINTRIN_H_INCLUDED.
18098         * config/i386/pconfigintrin.h: Ditto.
18099         * config/i386/waitpkgintrin.h: Ditto.
18100         * config/i386/immintrin.h: Add includes for sgxintrin.h,
18101         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
18102         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
18103         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
18104         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
18105         waitpkgintrin.h and cldemoteintrin.h.
18107 2018-06-04  Richard Biener  <rguenther@suse.de>
18109         PR tree-optimization/86038
18110         * tracer.c (find_best_successor): Check probability for
18111         being initialized, bail out if not.
18113 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
18115         PR target/86003
18116         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
18117         of bits to ignore when comparing architectures.
18119 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
18121         PR tree-optimization/69615
18122         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
18123         maximum or minimum of the type, try to merge it also as if
18124         range1 is + [-, x - 1] or + [x + 1, -].
18126         PR c++/86025
18127         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
18129 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
18131         PR tree-optimization/86034
18132         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
18133         the unsigned bitfield type in a bit insertion sequence if it does not
18134         have a larger precision than the bitfield size.
18135         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
18137 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
18139         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
18141 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18143         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
18144         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
18145         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
18146         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
18148 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18150         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
18151         Disable -fdelete-null-pointer-checks for ELF toolchain.
18153 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18154             Kito Cheng  <kito.cheng@gmail.com>
18156         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
18157         (nds32le-*-*, nds32be-*-*): Integrate checking process.
18158         (nds32*-*-*): Add glibc and uclibc conditions.
18159         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
18160         (TARGET_EXCEPT_UNWIND_INFO): Define.
18161         * config/nds32/elf.h: New file.
18162         * config/nds32/linux.h: New file.
18163         * config/nds32/nds32-elf.opt: New file.
18164         * config/nds32/nds32-linux.opt: New file.
18165         * config/nds32/nds32-fp-as-gp.c
18166         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
18167         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
18168         TARGET_LINUX_ABI.
18169         (nds32_asm_file_end): Ditto.
18170         (nds32_print_operand): Ditto.
18171         (nds32_insert_attributes): Ditto.
18172         (nds32_init_libfuncs): New function.
18173         (TARGET_HAVE_TLS): Define.
18174         (TARGET_INIT_LIBFUNCS): Define.
18175         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
18176         spec content.
18177         (TARGET_ELF): Apply different mcmodel setting.
18178         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
18179         been migrated into elf.h and linux.h files.
18180         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
18181         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
18182         (mcmodel): The content has been migrated into nds32-elf.opt and
18183         nds32-linux.opt files.
18184         * config/nds32/t-elf: New file.
18185         * config/nds32/t-linux: New file.
18187 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18188             Shiva Chen  <shiva0217@gmail.com>
18190         * config/nds32/constants.md (unspec_volatile_element): Add
18191         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
18192         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
18193         optimization.
18194         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
18195         (make_pass_nds32_fp_as_gp): Declare.
18196         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
18197         optmization pass.
18198         (nds32_asm_function_end_prologue): Remove unused asm output.
18199         (nds32_asm_function_begin_epilogue): Remove unused asm output.
18200         (nds32_asm_file_start): Output necessary fp_as_gp information.
18201         (nds32_option_override): Adjust register usage.
18202         (nds32_expand_prologue): Consider fp_as_gp situation.
18203         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
18204         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
18205         (epilogue): Ditto.
18206         (return): Ditto.
18207         (simple_return): Ditto.
18208         (omit_fp_begin): Output special directive for fp_as_gp.
18209         (omit_fp_end): Output special directive for fp_as_gp.
18210         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
18211         mforbid-fp-as-gp): New options.
18213 2018-06-01  Mark Wielaard  <mark@klomp.org>
18215         * dwarf2out.c (dwarf2out_finish): Remove generation of
18216         DW_AT_loclists_base.
18218 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
18220         * gimple-ssa-store-merging.c: Include gimple-fold.h.
18221         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
18222         (struct merged_store_group): Add bit_insertion field.
18223         (dump_char_array): Use standard hexadecimal format.
18224         (merged_store_group::merged_store_group): Set bit_insertion to false.
18225         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
18226         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
18227         also print the mask in the dump file.
18228         (pass_store_merging::gate): Minor tweak.
18229         (imm_store_chain_info::coalesce_immediate): Fix wrong association
18230         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
18231         stores with INTEGER_CST stores.
18232         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
18233         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
18234         and use it throughout.  Generate bit insertion sequences if need be.
18235         (pass_store_merging::process_store): Remove redundant condition.
18236         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
18238 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
18240         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
18241         the 128-bit floating point types.  Fix function comment.
18243 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18245         * config/aarch64/aarch64-simd.md
18246         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
18247         mnemonics.
18248         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
18249         mnemonics.
18251 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
18253         PR tree-optimization/85989
18254         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
18255         variable.
18256         (backprop::intersect_uses): Check it when deciding whether this
18257         is a backedge reference.
18258         (backprop::process_block): Add each phi to m_visited_phis
18259         after visiting it, then clear it at the end.
18261 2018-06-01  Richard Biener  <rguenther@suse.de>
18263         * tree-vectorizer.h (vect_dr_stmt): New function.
18264         (vect_get_load_cost): Adjust.
18265         (vect_get_store_cost): Likewise.
18266         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
18267         Use vect_dr_stmt instead of DR_SMTT.
18268         (vect_record_base_alignments): Likewise.
18269         (vect_calculate_target_alignment): Likewise.
18270         (vect_compute_data_ref_alignment): Likewise and make static.
18271         (vect_update_misalignment_for_peel): Likewise.
18272         (vect_verify_datarefs_alignment): Likewise.
18273         (vector_alignment_reachable_p): Likewise.
18274         (vect_get_data_access_cost): Likewise.  Pass down
18275         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
18276         (vect_get_peeling_costs_all_drs): Likewise.
18277         (vect_peeling_hash_get_lowest_cost): Likewise.
18278         (vect_enhance_data_refs_alignment): Likewise.
18279         (vect_find_same_alignment_drs): Likewise.
18280         (vect_analyze_data_refs_alignment): Likewise.
18281         (vect_analyze_group_access_1): Likewise.
18282         (vect_analyze_group_access): Likewise.
18283         (vect_analyze_data_ref_access): Likewise.
18284         (vect_analyze_data_ref_accesses): Likewise.
18285         (vect_vfa_segment_size): Likewise.
18286         (vect_small_gap_p): Likewise.
18287         (vectorizable_with_step_bound_p): Likewise.
18288         (vect_prune_runtime_alias_test_list): Likewise.
18289         (vect_analyze_data_refs): Likewise.
18290         (vect_supportable_dr_alignment): Likewise.
18291         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
18292         (vect_gen_prolog_loop_niters): Likewise.
18293         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
18294         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
18295         modify DR_STMT.
18296         (vect_recog_mask_conversion_pattern): Likewise.
18297         (vect_try_gather_scatter_pattern): Likewise.
18298         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
18299         to vect_get_store_cost.
18300         (vect_get_store_cost): Get stmt_info instead of DR.
18301         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
18302         (vect_get_load_cost): Get stmt_info instead of DR.
18304 2018-06-01  Richard Biener  <rguenther@suse.de>
18306         PR middle-end/86017
18307         * gimple-fold.c (var_decl_component_p): Also allow offsetted
18308         vars wrapped in MEM_REFs.
18310 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
18312         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
18313         Fix subreg tests so that we only return a choice between
18314         GENERAL_REGS and FP_REGS if the original classes included both.
18316 2018-06-01  Richard Biener  <rguenther@suse.de>
18318         PR ipa/85960
18319         * tree-ssa-structalias.c (get_function_part_constraint):
18320         Handle NULL fi->decl.
18321         (find_func_aliases_for_call): Properly handle indirect
18322         fi from direct call.
18323         (find_func_clobbers): Likewise.
18324         (ipa_pta_execute): Likewise.
18325         (create_variable_info_for): For functions that are ifunc_resolver
18326         resolve to a varinfo that contains the result of the resolver call.
18327         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
18328         aliases.
18330 2018-05-31  Michael Collison  <michael.collison@arm.com>
18332         * config/aarch64/aarch64.md:
18333         (*fix_to_zero_extenddfdi2): New pattern.
18334         * gcc.target/aarch64/fix_extend1.c: New testcase.
18336 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
18338         PR middle-end/78809
18339         PR middle-end/83026
18340         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
18341         and BUILT_IN_STRNCMP_EQ.
18342         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
18343         BUILT_IN_STRNCMP_EQ.
18344         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
18345         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
18346         (gimple_fold_builtin): Likewise.
18347         * tree-ssa-strlen.c (compute_string_length): New function.
18348         (determine_min_obsize): New function.
18349         (handle_builtin_string_cmp): New function to handle calls to
18350         string compare functions.
18351         (strlen_optimize_stmt): Add handling to builtin string compare
18352         calls.
18353         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
18354         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
18355         * tree.c (build_common_builtin_nodes): Add new defines of
18356         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
18358 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
18360         PR target/85984
18361         * bb-reorder.c (pass_partition_blocks::gate): Return false for
18362         functions with naked attribute.
18364 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
18366         * config/i386/sse.md (avx_vec_concat<mode>):
18367         Substitute concat_tg_mode mode attribute with xtg_mode.
18368         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
18369         (concat_tg_mode): Remove mode attribute.
18371 2018-05-31  Martin Sebor  <msebor@redhat.com>
18373         PR c/82063
18374         * calls.c (alloc_max_size): Correct a logic error/typo.
18375         Treat excessive arguments as infinite.  Warn for invalid arguments.
18376         * doc/invoke.texi (-Walloc-size-larger-than): Update.
18378 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
18380         PR target/85829
18381         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
18382         and movx for Haswell.
18384 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
18385             Cesar Philippidis  <cesar@codesourcery.com>
18387         PR middle-end/85879
18388         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
18389         when emitting error on private/firstprivate reductions.
18390         * omp-low.c (lower_omp_target): Avoid reference-type processing
18391         on pointers for firstprivate clause.
18393 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
18395         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
18396         (st1x2): Likewise.
18397         (st1x3): Likewise.
18398         * config/aarch64/aarch64-simd.md
18399         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
18400         (aarch64_ld1_x3_<mode>): Likewise
18401         (aarch64_st1x2<VALLDIF:mode>): Likewise
18402         (aarch64_st1_x2_<mode>): Likewise
18403         (aarch64_st1x3<VALLDIF:mode>): Likewise
18404         (aarch64_st1_x3_<mode>): Likewise
18405         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
18406         (vld1_s8_x3): Likewise.
18407         (vld1_u16_x3): Likewise.
18408         (vld1_s16_x3): Likewise.
18409         (vld1_u32_x3): Likewise.
18410         (vld1_s32_x3): Likewise.
18411         (vld1_u64_x3): Likewise.
18412         (vld1_s64_x3): Likewise.
18413         (vld1_f16_x3): Likewise.
18414         (vld1_f32_x3): Likewise.
18415         (vld1_f64_x3): Likewise.
18416         (vld1_p8_x3): Likewise.
18417         (vld1_p16_x3): Likewise.
18418         (vld1_p64_x3): Likewise.
18419         (vld1q_u8_x3): Likewise.
18420         (vld1q_s8_x3): Likewise.
18421         (vld1q_u16_x3): Likewise.
18422         (vld1q_s16_x3): Likewise.
18423         (vld1q_u32_x3): Likewise.
18424         (vld1q_s32_x3): Likewise.
18425         (vld1q_u64_x3): Likewise.
18426         (vld1q_s64_x3): Likewise.
18427         (vld1q_f16_x3): Likewise.
18428         (vld1q_f32_x3): Likewise.
18429         (vld1q_f64_x3): Likewise.
18430         (vld1q_p8_x3): Likewise.
18431         (vld1q_p16_x3): Likewise.
18432         (vld1q_p64_x3): Likewise.
18433         (vst1_s64_x2): Likewise.
18434         (vst1_u64_x2): Likewise.
18435         (vst1_f64_x2): Likewise.
18436         (vst1_s8_x2): Likewise.
18437         (vst1_p8_x2): Likewise.
18438         (vst1_s16_x2): Likewise.
18439         (vst1_p16_x2): Likewise.
18440         (vst1_s32_x2): Likewise.
18441         (vst1_u8_x2): Likewise.
18442         (vst1_u16_x2): Likewise.
18443         (vst1_u32_x2): Likewise.
18444         (vst1_f16_x2): Likewise.
18445         (vst1_f32_x2): Likewise.
18446         (vst1_p64_x2): Likewise.
18447         (vst1q_s8_x2): Likewise.
18448         (vst1q_p8_x2): Likewise.
18449         (vst1q_s16_x2): Likewise.
18450         (vst1q_p16_x2): Likewise.
18451         (vst1q_s32_x2): Likewise.
18452         (vst1q_s64_x2): Likewise.
18453         (vst1q_u8_x2): Likewise.
18454         (vst1q_u16_x2): Likewise.
18455         (vst1q_u32_x2): Likewise.
18456         (vst1q_u64_x2): Likewise.
18457         (vst1q_f16_x2): Likewise.
18458         (vst1q_f32_x2): Likewise.
18459         (vst1q_f64_x2): Likewise.
18460         (vst1q_p64_x2): Likewise.
18461         (vst1_s64_x3): Likewise.
18462         (vst1_u64_x3): Likewise.
18463         (vst1_f64_x3): Likewise.
18464         (vst1_s8_x3): Likewise.
18465         (vst1_p8_x3): Likewise.
18466         (vst1_s16_x3): Likewise.
18467         (vst1_p16_x3): Likewise.
18468         (vst1_s32_x3): Likewise.
18469         (vst1_u8_x3): Likewise.
18470         (vst1_u16_x3): Likewise.
18471         (vst1_u32_x3): Likewise.
18472         (vst1_f16_x3): Likewise.
18473         (vst1_f32_x3): Likewise.
18474         (vst1_p64_x3): Likewise.
18475         (vst1q_s8_x3): Likewise.
18476         (vst1q_p8_x3): Likewise.
18477         (vst1q_s16_x3): Likewise.
18478         (vst1q_p16_x3): Likewise.
18479         (vst1q_s32_x3): Likewise.
18480         (vst1q_s64_x3): Likewise.
18481         (vst1q_u8_x3): Likewise.
18482         (vst1q_u16_x3): Likewise.
18483         (vst1q_u32_x3): Likewise.
18484         (vst1q_u64_x3): Likewise.
18485         (vst1q_f16_x3): Likewise.
18486         (vst1q_f32_x3): Likewise.
18487         (vst1q_f64_x3): Likewise.
18488         (vst1q_p64_x3): Likewise.
18490 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18492         * config/msp430/msp430.c (msp430_output_labelref): Prepend
18493         user_label_prefix to name.
18495         * tree-core.h: Update comment about the format of NAME string
18496         passed to handler in attribute_spec.
18498         * config/msp430/msp430.md: Remove erroneous subreg expression from
18499         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
18500         zero_extend{q,h}isi2.
18502 2018-05-30  Borislav Petkov  <bp@suse.de>
18504         * doc/extend.texi: Document some architecture specific
18505         constraints and sort entries.
18507 2018-05-30  Martin Sebor  <msebor@redhat.com>
18509         PR middle-end/85369
18510         * builtins.c (expand_builtin_stpcpy_1): New function.
18511         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
18512         only if the former succeeds.
18514 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
18516         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
18517         in saphira.
18519 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
18521         * doc/invoke.texi (-flinker-output): Document
18523 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
18525         * passes.c (ipa_write_summaries): Only modify statements if body
18526         is in memory.
18527         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
18528         incrementally linking.
18529         (ipa_passes): Likewise.
18530         * lto-cgraph.c (lto_output_node): When incrementally linking do not
18531         pass down resolution info.
18532         * common.opt (flag_incremental_link): Update info.
18533         * gcc.c (plugin specs): Turn flinker-output=* to
18534         -plugin-opt=-linker-output-known
18535         * toplev.c (compile_file): Also cut compilation when doing incremental
18536         link.
18537         * flag-types. (enum lto_partition_model): Add
18538         LTO_LINKER_OUTPUT_NOLTOREL.
18539         (invoke.texi): Add -flinker-output docs.
18540         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
18541         link same way as WPA; do not stream in dead initializers.
18543         * dwarf2out.c (dwarf2out_die_ref_for_decl,
18544         darf2out_register_external_decl): Support incremental link.
18546 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
18548         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
18550 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
18552         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
18553         it down to simple_object_copy_lto_debug_sections.
18554         (run_gcc): Determine incremental LTO link time and configure
18555         lto1 into non-wpa mode, disable renaming of debug sections.
18557 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18559         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
18560         descriptions of various incorrectly documented functions.
18562 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18564         Revert:
18565         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18566         address check not strict.
18568 2018-05-30  Richard Biener  <rguenther@suse.de>
18570         PR tree-optimization/85964
18571         * tracer.c (better_p): Drop initialized count check, we only
18572         call the function with initialized counts now.
18573         (find_best_successor): Do find a best edge if one
18574         has uninitialized count.
18575         (find_best_predecessor): Likewise.  Do BB frequency check only
18576         if count is initialized.
18578 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
18580         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
18581         (aarch64_ldrstr_offset_compare): New.
18582         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
18583         load/store orderings.
18584         (aarch64_gen_adjusted_ldpstp): Likewise.
18586 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
18588         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
18589         Check for subset of GENERAL_REGS and FP_REGS.
18590         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
18591         r=w alternative.
18593 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
18595         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
18596         and wi::to_poly_offset.  Add the current offset and then check
18597         whether the sum fits, rather than using an unchecked addition of
18598         a checked term.  Check for a shwi rather than a uhwi.
18599         * expr.c (get_bit_range): Use tree_to_poly_uint64.
18600         (store_constructor): Use poly_int_tree_p.
18601         (expand_expr_real_1): Likewise.
18602         * function.c (assign_temp): Likewise.
18603         * fold-const.c (const_binop): Use poly_int_tree_p and
18604         wi::to_poly_offset.
18605         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
18606         division.
18607         * ipa-icf-gimple.c (func_checker::compare_operand): Use
18608         to_poly_offset for MEM offsets.
18609         * ipa-icf.c (sem_variable::equals): Likewise.
18610         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
18611         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
18612         wi::to_poly_offset for BIT_FIELD_REF offsets.
18613         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
18614         wi::to_poly_offset.
18615         * var-tracking.c (emit_note_insn_var_location): Use
18616         tree_to_poly_uint64.
18618 2018-05-29  Jim Wilson  <jimw@sifive.com>
18620         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
18622 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
18624         PR target/85950
18625         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
18626         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
18627         sequence.
18628         (sse4_1_round<mode>2): Use nonimmediate_operand
18629         for operand 1 predicate.
18631 2018-05-29  Martin Sebor  <msebor@redhat.com>
18632             Richard Biener  <rguenther@suse.de>
18634         PR testsuite/85888
18635         * calls.c (get_size_range): Call determine_value_range instead
18636         of get_value_range..
18637         * tree-vrp.h (determine_value_range): Declared new function.
18638         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
18640 2018-05-29  Richard Biener  <rguenther@suse.de>
18642         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
18643         sure to use non-pattern stmts for get_earlier_stmt arguments.
18644         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
18645         called on pattern stmts.
18646         (get_later_stmt): Likewise.
18648 2018-05-29  Martin Liska  <mliska@suse.cz>
18650         PR gcov-profile/85759
18651         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
18652         env variables.
18654 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
18656         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
18657         VEC_UNPACK_*_EXPR.
18658         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
18659         VEC_PACK_*_EXPR.
18661         PR target/85918
18662         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18663         VEC_PACK_FLOAT_EXPR): New tree codes.
18664         * tree-pretty-print.c (op_code_prio): Handle
18665         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18666         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18667         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18668         * tree-inline.c (estimate_operator_cost): Likewise.
18669         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
18670         * fold-const.c (const_binop): Likewise.
18671         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
18672         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18673         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18674         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
18675         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18676         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18677         * expr.c (expand_expr_real_2): Likewise.
18678         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
18679         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
18680         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
18681         optabs.
18682         * optabs.c (expand_widen_pattern_expr): For
18683         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18684         sign from result type rather than operand's type.
18685         (expand_binop_directly): For vec_packu_float_optab and
18686         vec_packs_float_optab allow result type to be different from operand's
18687         type.
18688         * optabs-tree.c (optab_for_tree_code): Handle
18689         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18690         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
18691         * tree-vect-generic.c (expand_vector_operations_1):  Handle
18692         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18693         VEC_PACK_FLOAT_EXPR.
18694         * tree-vect-stmts.c (supportable_widening_operation): Handle
18695         FIX_TRUNC_EXPR.
18696         (supportable_narrowing_operation): Handle FLOAT_EXPR.
18697         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18698         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18699         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
18700         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18701         mode attributes.
18702         (vec_pack<floatprefix>_float_<mode>): New expander.
18703         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18704         attributes.
18705         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18706         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18707         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18708         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18709         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18710         Document.
18711         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18712         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18713         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18714         VEC_PACK_FLOAT_EXPR): Document.
18716 2018-05-29  Richard Biener  <rguenther@suse.de>
18718         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18719         member.
18720         (stmt_vec_info_vec): Make pointer.
18721         (init_stmt_vec_info_vec): Remove.
18722         (free_stmt_vec_info_vec): Likewise.
18723         (set_stmt_vec_info_vec): New function.
18724         (free_stmt_vec_infos): Likewise.
18725         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18726         (set_vinfo_for_stmt): Likewise.
18727         (get_earlier_stmt): Likewise.
18728         (get_later_stmt): Likewise.
18729         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18730         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18731         (vec_info::~vec_info): Free stmt_vec_infos.
18732         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18733         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18734         (pass_slp_vectorize::execute): Likewise.
18735         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18736         (free_stmt_vec_info_vec): Likewise.
18737         (set_stmt_vec_info_vec): New function.
18738         (free_stmt_vec_infos): Likewise.
18739         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18740         the global stmt_vec_info_vec.
18741         * tree-parloops.c (gather_scalar_reductions): Use
18742         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18743         vector.
18745 2018-05-29  Richard Biener  <rguenther@suse.de>
18747         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18749 2018-05-29  Martin Liska  <mliska@suse.cz>
18750             David Malcolm  <dmalcolm@redhat.com>
18752         * vec.c (test_reverse): New.
18753         (vec_c_tests): Add new test.
18754         * vec.h (vl_ptr>::reverse): New function.
18756 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
18758         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18760         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18761         and later.
18763 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18765         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18767 2018-05-28  Richard Biener  <rguenther@suse.de>
18769         PR tree-optimization/85933
18770         * tree-vect-data-refs.c (vect_record_base_alignments): Only
18771         look at stmts marked as vectorizable.
18773 2018-05-28  Richard Biener  <rguenther@suse.de>
18775         PR tree-optimization/85934
18776         * tree-vect-generic.c (expand_vector_operations_1): Hoist
18777         vector boolean check before scalar optimization.
18779 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
18781         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18782         for armv5te.
18784 2018-05-28  Mark Wielaard  <mark@klomp.org>
18786         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18787         if it is an expression containing a minus sign.
18789 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
18791         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18793 2018-05-27  Paul Koning  <ni1d@arrl.net>
18795         * config/pdp11/pdp11.md (truncsihi2): Remove.
18797 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
18798             Chung-Ju Wu  <jasonwucj@gmail.com>
18800         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18801         implementation.
18802         (unaligned_store_dw): Ditto.
18803         * config/nds32/nds32-memory-manipulation.c
18804         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18805         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18806         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18807         (emit_setmem_word_loop): Rename to ...
18808         (emit_setmem_doubleword_loop): ... this.
18809         (nds32_gen_dup_4_byte_to_word_value): New function.
18810         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18811         (nds32_expand_setmem_loop): Refine implementation.
18812         (nds32_expand_setmem_loop_v3m): Ditto.
18813         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18814         pattern.
18816 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18818         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18820 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18822         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18823         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18824         machine->attr_no_prologue_p.
18825         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18826         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18827         (nds32_expand_epilogue): Consider attr_naked_p.
18828         (nds32_expand_epilogue_v3pop): Likewise.
18829         (nds32_can_use_return_insn): Likewise.
18830         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18831         attr_no_prologue_p fields.
18832         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18834 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18836         PR target/85918
18837         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18838         attributes.
18839         * config/i386/sse.md
18840         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18841         Rename to ...
18842         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18843         ... this.
18844         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18845         Rename to ...
18846         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18847         ... this.
18848         (*<floatsuffix>floatv2div2sf2): Rename to ...
18849         (*float<floatunssuffix>v2div2sf2): ... this.
18850         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18851         (float<floatunssuffix>v2div2sf2_mask): ... this.
18852         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18853         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18854         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18855         to ...
18856         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18857         ... this.
18858         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18859         Rename to ...
18860         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18861         ... this.
18862         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18863         Rename to ...
18864         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18865         ... this.
18866         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18867         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18868         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18869         gen_ufix_truncv8dfv8si2.
18870         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18871         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18872         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18873         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18874         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18875         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18876         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18877         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18879 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18881         PR target/85900
18882         PR target/85345
18883         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18885 2018-05-25  Jim Wilson  <jimw@sifive.com>
18887         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18888         * config/riscv/riscv.c (struct machine_function): Add
18889         interrupt_handler_p and attribute_checked_p fields.
18890         (riscv_attribute_table): Add interrupt.
18891         (riscv_interrupt_type_p): New.
18892         (riscv_save_reg_p): Save extra regs for interrupt handler.
18893         (riscv_use_save_libcall): Return false  for interrupt handler.
18894         (riscv_first_stack_step): Add forward declaration.
18895         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18896         for interrupt handler with large frame.  Use it for saved reg list.
18897         (riscv_expand_prologue): Move flag_stack_usage_info support to
18898         eliminate duplication.
18899         (riscv_expand_epilogue): Generate mret for interrupt handler.
18900         (riscv_epilogue_uses): New.
18901         (riscv_can_use_return_insn): Return false for interrupt handler.
18902         (riscv_function_ok_for_sibcall): Likewise.
18903         (riscv_set_current_function): Add interrupt handler support.
18904         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18905         * config/riscv/riscv.md (UNSPECV_MRET): New.
18906         (GP_REGNUM): New.
18907         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18908         (riscv_mret): New.
18909         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18911 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18913         PR tree-optimization/85712
18914         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18915         this candidate has already been replaced in-situ by a copy.
18917 2018-05-25  Jason Merrill  <jason@redhat.com>
18919         PR c++/80485 - inline function non-zero address.
18920         * symtab.c (nonzero_address): Check DECL_COMDAT.
18922 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18924         PR target/83628
18925         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18926         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18927         extension of SImode operation.  Use const123_operand predicate.
18928         (*saddsi_1): Remove.
18929         (*saddl_se_1): Ditto.
18930         (*ssubsi_1): Ditto.
18931         (*ssubl_se_1): Ditto.
18932         * config/alpha/predicates.md (const123_operand): New predicate.
18933         * config/alpha/constraints.md (P): Use IN_RANGE.
18935 2018-05-25  Richard Biener  <rguenther@suse.de>
18937         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18938         defaulted to true.
18939         (ref_maybe_used_by_stmt_p): Likewise.
18940         (stmt_may_clobber_ref_p): Likewise.
18941         (stmt_may_clobber_ref_p_1): Likewise.
18942         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18943         and pass it along.
18944         (ref_maybe_used_by_stmt_p): Likewise.
18945         (stmt_may_clobber_ref_p): Likewise.
18946         (stmt_may_clobber_ref_p_1): Likewise.
18947         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18948         the alias oracle to disambiguate DRs with stmts DR analysis
18949         couldn't handle.
18950         (vect_analyze_data_refs): Do not give up on not analyzable
18951         DRs for BB vectorization.  Remove code truncating the dataref
18952         vector.
18954 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18956         PR target/85832
18957         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18958         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18959         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18961 2018-05-25  Richard Biener  <rguenther@suse.de>
18963         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18964         function, combining stmt data ref gathering and fatal analysis
18965         parts.
18966         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18967         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18968         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18969         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18970         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18972 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18974         PR tree-optimization/85720
18975         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18976         SCC if all partitions are builtins.
18977         (version_loop_by_alias_check): New parameter.  Generate cancelable
18978         runtime alias check if all partitions are builtins.
18979         (distribute_loop): Update call to above function.
18981 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18983         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18984         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18985         (parm_default_def_partition_arg): Ditto.
18986         (set_parm_default_def_partition): Ditto.
18987         (get_parm_default_def_partitions): Ditto and make it static.
18988         (get_undefined_value_partitions): Ditto and make it static.
18989         (remove_ssa_form): Refactor call to init_var_map here.
18990         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18991         computation for loop region.
18992         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18993         (register_default_def): Delete.
18994         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18995         (parm_default_def_partition_arg): Ditto.
18996         (set_parm_default_def_partition): Ditto.
18997         (get_parm_default_def_partitions): Ditto and make it static.
18998         (get_undefined_value_partitions): Ditto and make it static.
18999         (coalesce_with_default, coalesce_with_default): Update comment.
19000         (create_coalesce_list_for_region): New func factored out from
19001         create_outofssa_var_map.
19002         (populate_coalesce_list_for_outofssa): New func factored out from
19003         create_outofssa_var_map and coalesce_ssa_name.
19004         (create_outofssa_var_map): Delete.
19005         (coalesce_ssa_name): Refactor to support live range computation.
19006         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
19007         (get_parm_default_def_partitions): Delete.
19008         (get_undefined_value_partitions): Ditto.
19009         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
19010         computation for loop region.
19011         (new_tree_live_info, loe_visit_block): Ditto.
19012         (live_worklist, set_var_live_on_entry): Ditto.
19013         (calculate_live_on_exit, verify_live_on_entry): Ditto.
19014         * tree-ssa-live.h (struct _var_map): New fields.
19015         (init_var_map): Change decl.
19016         (region_contains_p): New.
19018 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
19020         * tree-ssa-live.h (live_merge_and_clear): Delete.
19022 2018-05-25  Richard Biener  <rguenther@suse.de>
19024         PR c++/85912
19025         * tree-dump.c (dequeue_and_dump): Remove access to removed
19026         operand 2 of a SWITCH_EXPR.
19028 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19030         * doc/sourcebuild.texi (vect_double_cond_arith): Include
19031         multiplication and division.
19032         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
19033         (cond_udiv@var{m}, cond_umod@var{m}): Document.
19034         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
19035         (cond_udiv_optab, cond_umod_optab): New optabs.
19036         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
19037         (IFN_COND_RDIV): New internal functions.
19038         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
19039         TRUNC_MOD_EXPR and RDIV_EXPR.
19040         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
19041         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
19042         New unspecs.
19043         (SVE_INT_BINARY): Include mult.
19044         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
19045         (optab, sve_int_op): Handle mult.
19046         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
19047         UNSPEC_COND_DIV.
19048         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
19049         for SVE_INT_BINARY_SD.
19051 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19053         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
19054         (optab, sve_int_op): Handle div and udiv.
19055         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
19056         for SVE_INT_BINARY_SD.
19057         (*<optab><mode>3): New insn for the same.
19059 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19061         * tree-vect-patterns.c: Include predict.h.
19062         (vect_recog_divmod_pattern): Restrict check for division support
19063         to when optimizing for size.
19065 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19067         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
19068         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
19069         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
19070         (gimple_match_op::set_op): Likewise.
19071         (gimple_resimplify4): Declare.
19072         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
19073         (expr::gen_transform): Likewise.
19074         (decision_tree::gen): Generate a simplification routine for 4 operands.
19075         * gimple-match-head.c (gimple_simplify): Add an overload for
19076         4 operands.  In the top-level function, handle up to 4 call
19077         arguments and call gimple_resimplify4.
19078         (gimple_resimplify4): New function.
19079         (build_call_internal): Pass a fourth operand.
19080         (maybe_push_to_seq): Likewise.
19081         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
19082         Fold VEC_COND_EXPRs of an operation and a default value into
19083         an IFN_COND_* function if possible.
19084         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
19085         New unspecs.
19086         (SVE_COND_FP_BINARY): Include them.
19087         (optab, sve_fp_op): Handle them.
19088         (SVE_INT_BINARY_REV): New code iterator.
19089         (SVE_COND_FP_BINARY_REV): New int iterator.
19090         (commutative): New int attribute.
19091         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
19092         Declare.
19093         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
19094         function.
19095         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
19096         (*cond_<optab><mode>): New patterns for reversed operands.
19098 2018-05-25  Richard Biener  <rguenther@suse.de>
19100         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
19101         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
19102         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
19103         (STMT_VINFO_GROUPED_ACCESS): Adjust.
19104         * tree-vect-data-refs.c (everywhere): Adjust users.
19105         * tree-vect-loop.c (everywhere): Likewise.
19106         * tree-vect-slp.c (everywhere): Likewise.
19107         * tree-vect-stmts.c (everywhere): Likewise.
19108         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
19110 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19112         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
19113         Rename to...
19114         (gcc_cv_as_section_exclude): ... this.
19115         Try Solaris as #exclude syntax.
19116         * configure: Regenerate.
19117         * config.in: Regenerate.
19118         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
19119         SECTION_EXCLUDE.
19120         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
19121         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
19123         * varasm.c (default_elf_asm_named_section): Don't check if
19124         HAVE_GAS_SECTION_EXCLUDE is defined.
19126 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19128         * doc/md.texi: Update the documentation of the cond_* optabs
19129         to mention the new final operand.  Fix GET_MODE_NUNITS call.
19130         Describe the scalar case too.
19131         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
19132         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
19133         instead of 2.
19134         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
19135         (get_conditional_internal_fn): Update comment.
19136         * tree-vect-loop.c (vectorizable_reduction): Pass the original
19137         accumulator value as a final argument to conditional functions.
19138         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
19139         a define_expand and add an "else" operand.  Assert for now that
19140         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
19141         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
19142         (*cond_<optab><mode>): New patterns.
19143         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
19144         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19145         (UNSPEC_COND_EOR): Delete.
19146         (optab): Remove associated mappings.
19147         (SVE_INT_BINARY): New code iterator.
19148         (sve_int_op): Remove int attribute and add "minus" to the code
19149         attribute.
19150         (SVE_COND_INT_OP): Delete.
19151         (SVE_COND_FP_OP): Rename to...
19152         (SVE_COND_FP_BINARY): ...this.
19154 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19156         * optabs.c (can_reuse_operands_p): New function.
19157         (maybe_legitimize_operands): Try to reuse the results for
19158         earlier operands.
19160 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
19162         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
19163         Add {q} suffix to insn mnemonic.
19165 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19167         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
19168         (msp430_warn_func_return): New.
19170 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
19172         * fold-const.c (tree_nonzero_bits): New function.
19173         * fold-const.h (tree_nonzero_bits): Likewise.
19174         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
19175         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
19177 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
19179         PR target/85900
19180         PR target/85345
19181         * varasm.c (assemble_alias): Check ifunc_resolver only on
19182         FUNCTION_DECL.
19184 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
19186         PR target/85903
19187         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
19188         when memory input operand is handled.
19190 2018-05-24  Luis Machado  <luis.machado@linaro.org>
19192         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
19193         global.
19194         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
19196 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
19198         * match.pd: Delay FMA folds until after vectorization.
19200 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
19202         PR target/83009
19203         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
19204         address check not strict.
19206 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
19208         * gimple-match.h (gimple_match_op): New class.
19209         (mprts_hook): Replace parameters with a gimple_match_op *.
19210         (maybe_build_generic_op): Likewise.
19211         (gimple_simplified_result_is_gimple_val): Replace parameters with
19212         a const gimple_match_op *.
19213         (gimple_simplify): Replace code_helper * and tree * parameters with
19214         a gimple_match_op * parameter.
19215         (gimple_resimplify1): Replace code_helper *, tree and tree *
19216         parameters with a gimple_match_op * parameter.
19217         (gimple_resimplify2): Likewise.
19218         (gimple_resimplify3): Likewise.
19219         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
19220         parameters with a gimple_match_op * parameter.
19221         * gimple-match-head.c (gimple_simplify): Change prototypes of
19222         auto-generated functions to take a gimple_match_op * instead of
19223         separate code_helper * and tree * parameters.  Make the same
19224         change in the top-level overload and update calls to the
19225         gimple_resimplify routines.  Update calls to the auto-generated
19226         functions and to maybe_push_res_to_seq in the publicly-facing
19227         operation-specific gimple_simplify overloads.
19228         (gimple_match_op::MAX_NUM_OPS): Define.
19229         (gimple_resimplify1): Replace rcode and ops with a single res_op
19230         parameter.  Update call to gimple_simplify.
19231         (gimple_resimplify2): Likewise.
19232         (gimple_resimplify3): Likewise.
19233         (mprts_hook): Replace parameters with a gimple_match_op *.
19234         (maybe_build_generic_op): Likewise.
19235         (build_call_internal): Replace type, nargs and ops with
19236         a gimple_match_op *.
19237         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
19238         with a single gimple_match_op *.  Update calls to mprts_hook,
19239         build_call_internal and gimple_simplified_result_is_gimple_val.
19240         Factor out code that is common to the tree_code and combined_fn cases.
19241         * genmatch.c (expr::gen_transform): Replace tem_code and
19242         tem_ops with a gimple_match_op called tem_op.  Update calls
19243         to the gimple_resimplify functions and maybe_push_res_to_seq.
19244         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
19245         res_ops.  Update call to the gimple_resimplify functions.
19246         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
19247         (decision_tree::gen): Make the functions take a gimple_match_op *
19248         called res_op instead of separate res_code and res_ops parameters.
19249         Update call accordingly.
19250         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
19251         and ops with a single res_op parameter.  Update calls to
19252         maybe_build_generic_op and maybe_push_res_to_seq.
19253         (fold_stmt_1): Update calls to gimple_simplify and
19254         replace_stmt_with_simplification.
19255         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
19256         and gimple_simplified_result_is_gimple_val.
19257         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
19258         gimple_simplify.
19259         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
19260         with a gimple_match_op *.
19261         (vn_nary_build_or_lookup): Likewise.  Update call to
19262         vn_nary_build_or_lookup_1.
19263         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
19264         gimple_match_op *.  Update calls to the gimple_resimplify routines
19265         and to gimple_simplified_result_is_gimple_val.
19266         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
19267         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
19268         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
19269         (visit_nary_op): Likewise.
19270         (visit_reference_op_load): Likewise.
19272 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19274         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
19275         modifier for printing the step amount.
19277 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
19279         PR target/78849
19280         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
19281         types.
19283 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
19285         * doc/sourcebuild.texi (Endianness): New subsubsection.
19287 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19289         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19290         <prefetch_dynamic_strides>: New const bool field.
19291         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19292         prefetch_dynamic_strides.
19293         (exynosm1_prefetch_tune): Likewise.
19294         (thunderxt88_prefetch_tune): Likewise.
19295         (thunderx_prefetch_tune): Likewise.
19296         (thunderx2t99_prefetch_tune): Likewise.
19297         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
19298         false.
19299         (aarch64_override_options_internal): Update to set
19300         PARAM_PREFETCH_DYNAMIC_STRIDES.
19301         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19302         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19303         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19304         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19305         prefetch-dynamic-strides setting.
19307 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19309         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19310         <minimum_stride>: New const int field.
19311         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19312         minimum_stride field defaulting to -1.
19313         (exynosm1_prefetch_tune): Likewise.
19314         (thunderxt88_prefetch_tune): Likewise.
19315         (thunderx_prefetch_tune): Likewise.
19316         (thunderx2t99_prefetch_tune): Likewise.
19317         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
19318         <default_opt_level>: Set to 3.
19319         (aarch64_override_options_internal): Update to set
19320         PARAM_PREFETCH_MINIMUM_STRIDE.
19321         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19322         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19323         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19324         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19325         stride is constant and is below the minimum stride threshold.
19327 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19329         * config/arm/arm-cpus.in (mode26): Delete.
19330         (armv4): Delete mode26 reference.
19331         * config/arm/arm.c (arm_configure_build_target): Delete use of
19332         isa_bit_mode26.
19334 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
19336         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
19337         New insn pattern.
19338         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
19339         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
19340         for non-SSE modes.
19341         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
19342         (floatunsdidf2): Ditto.
19344 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
19346         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
19347         (fixuns_trunc<mode>si2_avx512f): Ditto.
19348         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
19349         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
19350         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
19352 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
19354         PR rtl-optimization/79985
19355         * df-scan.c (df_insn_refs_collect): Remove special case for
19356         global registers and asm statements.
19358 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
19360         * extend.texi (Global Register Variables): Rewrite the bullet list.
19361         Note that the register is available for allocation. Note that access
19362         via inline asm must use constraints. Add note about async-signal
19363         handlers. Remove paragraph about automagic register selection.
19365 2018-05-23  Richard Biener  <rguenther@suse.de>
19367         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
19368         of fixed offset from memset VN.
19370 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
19372         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
19373         first_interp field.
19374         (alloc_cand_and_find_basis): Initialize first_interp field.
19375         (slsr_process_mul): Modify first_interp field.
19376         (slsr_process_add): Likewise.
19377         (slsr_process_cast): Modify first_interp field for each new
19378         interpretation.
19379         (slsr_process_copy): Likewise.
19380         (dump_candidate): Dump first_interp field.
19381         (replace_mult_candidate): Process all interpretations, not just
19382         subsequent ones.
19383         (replace_rhs_if_not_dup): Likewise.
19384         (replace_one_candidate): Likewise.
19386 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
19388         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
19389         Add new boolean.
19390         (aarch64_needs_frame_chain): New function.
19391         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
19393 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
19395         PR target/84882
19396         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
19397         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
19398         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
19399         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
19400         as true for strict-align.
19401         (aarch64_can_inline_p): Perform checks even when callee has no
19402         attributes to check for strict alignment.
19403         * doc/extend.texi (AArch64 Function Attributes): Document
19404         no-strict-align.
19405         * doc/invoke.texi: (AArch64 Options): Likewise.
19407 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
19409         PR tree-optimization/85853
19410         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
19411         the handling of the root of the node to...
19412         (vect_slp_analyze_node_operations_1): ...this new function,
19413         and run the whole thing with the child nodes' def types
19414         set according to their SLP node's def type.
19416 2018-05-23  Richard Biener  <rguenther@suse.de>
19418         PR middle-end/85874
19419         * tree-data-ref.c (create_runtime_alias_checks): Defer
19420         and ignore overflow warnings.
19422 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
19424         PR tree-optimization/85822
19425         * tree-vrp.c (is_masked_range_test): Fix handling of negative
19426         constants.
19428 2018-05-23  Richard Biener  <rguenther@suse.de>
19430         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
19431         memset constants via native_interpret_expr.
19433 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
19435         PR target/85345
19436         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
19437         attribute.
19438         (cgraph_node::create_alias): Likewise.
19439         (cgraph_node::get_availability): Check ifunc_resolver instead
19440         of looking up ifunc attribute.
19441         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
19442         * varasm.c (do_assemble_alias): Likewise.
19443         (assemble_alias): Likewise.
19444         (default_binds_local_p_3): Likewise.
19445         * cgraph.h (cgraph_node): Add ifunc_resolver.
19446         (cgraph_node::only_called_directly_or_aliased_p): Return false
19447         for IFUNC resolver.
19448         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
19449         attribute.
19450         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
19451         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
19452         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
19453         instead of looking up ifunc attribute.
19455 2018-05-22  Luis Machado  <luis.machado@linaro.org>
19457         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
19459 2018-05-22  Martin Sebor  <msebor@redhat.com>
19461         PR middle-end/85359
19462         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
19463         only when expasion succeeds.
19464         (expand_builtin_strcmp): Same.
19465         (expand_builtin_strncmp): Same.
19467 2018-05-22  Martin Sebor  <msebor@redhat.com>
19469         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
19471 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
19472             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19474         * config/aarch64/aarch64-ldpstp.md: Replace uses of
19475         aarch64_mem_pair_operand with memory_operand and delete operand swapping
19476         code.
19477         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
19478         Add check for legitimate_address.
19479         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
19480         (aarch64_swap_ldrstr_operands): New.
19481         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
19482         Define prototype.
19484 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
19485             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19487         * config/aarch64/aarch64.md: New patterns to generate stp
19488         and ldp.
19489         (store_pair_sw, store_pair_dw): New patterns to generate stp for
19490         single words and double words.
19491         (load_pair_sw, load_pair_dw): Likewise.
19492         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
19493         Delete.
19494         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
19495         Delete.
19496         * config/aarch64/aarch64-ldpstp.md: Modify peephole
19497         for different mode ldpstp and add peephole for merged zero stores.
19498         Likewise for loads.
19499         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
19500         Add size check.
19501         (aarch64_gen_store_pair): Rename calls to match new patterns.
19502         (aarch64_gen_load_pair): Rename calls to match new patterns.
19503         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
19504         (load_pair<DREG:mode><DREG2:mode>): ... This.
19505         (store_pair<mode>): Rename to...
19506         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
19507         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
19508         New mode iterators.
19509         (V_INT_EQUIV): Handle SImode.
19510         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
19511         New predicate.
19513 2018-05-22  Martin Sebor  <msebor@redhat.com>
19515         PR c/85623
19516         * calls.c (maybe_warn_nonstring_arg): Use string length to set
19517         or ajust the presumed bound on an operation to avoid unnecessary
19518         warnings.
19520 2018-05-22  Martin Sebor  <msebor@redhat.com>
19522         PR tree-optimization/85826
19523         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
19524         assuming that a DECL necesarily has a constant size.
19526 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
19528         PR middle-end/85862
19529         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
19531 2018-05-22  Richard Biener  <rguenther@suse.de>
19533         PR tree-optimization/85834
19534         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
19535         non-constant and non-zero memset arguments.
19537 2018-05-22  Martin Liska  <mliska@suse.cz>
19539         PR ipa/85607
19540         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
19542 2018-05-22  Richard Biener  <rguenther@suse.de>
19544         PR tree-optimization/85863
19545         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
19546         comparisons when vectype is specified.
19547         (vectorizable_condition): Do not specify vectype for
19548         vect_is_simple_cond when SLP vectorizing.
19550 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
19552         PR target/85657
19553         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
19554         define __ibm128 as long double.
19555         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
19556         as a distinct type when IEEE 128-bit support is enabled.
19557         (init_float128_ieee): Fix up conversions between IFmode and IEEE
19558         128-bit types to use the correct functions.
19559         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
19560         convert between 128-bit floating point types that have different
19561         modes but the same representation, instead of using gen_lowpart to
19562         makean alias.
19563         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
19564         KFmode.
19565         (IFKF_reg): New attributes to give the register constraints for
19566         IFmode and KFmode.
19567         (extend<mode>tf2_internal): New insns to mark an explicit
19568         conversion between 128-bit floating point types that have a
19569         different mode but share the same representation.
19571 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
19573         PR tree-optimization/85814
19574         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
19575         a null return from get_strinfo when unsharing the next
19576         strinfo in the chain.
19578 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
19580         PR gcc/84923
19581         * varasm.c (weak_finish): Clean up weak_decls.
19583 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19585         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
19586         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
19587         UNSPEC_UADALP values.
19588         * config/aarch64/iterators.md (ABAL): New int iterator.
19589         (ABDL2): Likewise.
19590         (ADALP): Likewise.
19591         (sur): Add mappings for the above.
19592         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
19593         New define_insn.
19594         (aarch64_<sur>abal<mode>_4): Likewise.
19595         (aarch64_<sur>adalp<mode>_3): Likewise.
19596         (<sur>sadv16qi): New define_expand.
19598 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
19600         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
19601         (*movdf_internal): Ditto.
19602         (*rcpsf2_sse): Ditto.
19603         (*rsqrtsf2_sse): Ditto.
19604         (*sqrt<mode>2_sse): Ditto.
19606 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
19608         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
19609         eor3q<mode>4.
19610         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
19611         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
19612         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
19613         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
19614         vbcaxq_s64): New.
19615         * config/aarch64/arm_neon.h: Likewise.
19616         * config/aarch64/iterators.md (VQ_I): New.
19618 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
19620         * config.gcc: Add arc/t-multilib-linux to tmake_file for
19621         arc*-*-linux*.
19622         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
19623         MULTILIB_DIRNAMES
19625 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19627         * config/nds32/constraints.md (S): New constraint.
19628         * config/nds32/nds32.md (call_internal): Use constraint S.
19629         (call_value_internal): Likewise.
19630         (sibcall_internal): Likewise.
19631         (sibcall_value_internal): Likewise.
19633 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
19634             Chung-Ju Wu  <jasonwucj@gmail.com>
19636         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
19637         into consideration.
19639 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
19640             Chung-Ju Wu  <jasonwucj@gmail.com>
19642         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
19643         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
19644         (nds32_rtx_costs_impl): Simplify.
19645         (nds32_address_cost_impl): Simplify.
19646         (nds32_init_rtx_costs): New function.
19647         (nds32_rtx_costs_speed_prefer): Likewise.
19648         (nds32_rtx_costs_size_prefer): Likewise.
19649         (nds32_address_cost_speed_prefer): Likewise.
19650         (nds32_address_cost_speed_fwprop): Likewise.
19651         (nds32_address_cost_size_prefer): Likewise.
19652         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
19653         * config/nds32/nds32.c (nds32_option_override): Use
19654         nds32_init_rtx_costs function.
19656 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19658         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
19659         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
19660         (TARGET_PIPELINE_N8): Likewise.
19661         (TARGET_PIPELINE_N10): Likewise.
19662         (TARGET_PIPELINE_N13): Likewise.
19663         (TARGET_PIPELINE_GRAYWOLF): Likewise.
19665 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
19667         * config/nds32/nds32-fpu.md: Update copyright year.
19669 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19671         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
19673 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19675         * config/nds32/nds32.c
19676         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
19677         * config/nds32/nds32.opt (minline-asm-r15): New option.
19679 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19681         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19682         MASK_HW_ABS.
19683         * config/nds32/nds32.md (abssi2): New pattern.
19685 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
19687         * config/i386/i386.md (rex64namesuffix): New mode attribute.
19688         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19689         Merge insn pattern from sse_cvtsi2ss<round_name> and
19690         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19691         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19692         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19693         using SWI48 mode iterator.
19694         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19695         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19696         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19697         pattern from sse_cvttss2si<round_saeonly_name>
19698         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19699         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19700         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19701         using SWI48 mode iterator.
19702         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19703         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19704         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19705         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19706         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19707         using SWI48 mode iterator.
19708         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19709         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19710         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19711         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19712         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19713         SWI48 mode iterator.
19714         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19715         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19716         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19717         pattern from sse_cvttsd2si<round_saeonly_name>
19718         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19720 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19722         * config/nds32/nds32-md-auxiliary.c
19723         (nds32_valid_smw_lwm_base_p): Refine.
19724         (nds32_output_smw_single_word): Refine.
19725         (nds32_output_smw_double_word): New.
19726         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19728 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19730         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19731         (nds32_output_stack_pop): Refine.
19732         (nds32_expand_unaligned_load): Refine.
19733         (nds32_expand_unaligned_store): Refine.
19735 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19736             Chung-Ju Wu  <jasonwucj@gmail.com>
19738         * config/nds32/constants.md: Add TP_REGNUM constant.
19739         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19740         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19741         UNSPEC_ADD32.
19742         * config/nds32/nds32-doubleword.md: Consider flag_pic.
19743         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19744         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19745         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19746         and PIC code generation.
19747         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19748         code generation.
19749         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19750         optimization.
19751         * config/nds32/nds32.md: Support TLS and PIC.
19752         * config/nds32/nds32.c: Support TLS and PIC.
19753         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19754         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19755         predicate.
19757 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19759         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19760         mode with E_ prefix.
19762 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19763             Chung-Ju Wu  <jasonwucj@gmail.com>
19765         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19766         * config/nds32/nds32-md-auxiliary.c
19767         (symbolic_reference_mentioned_p): New.
19768         (nds32_legitimize_ict_address): New.
19769         (nds32_expand_ict_move): New.
19770         (nds32_indirect_call_referenced_p): New.
19771         (nds32_symbol_binds_local_p): Delete.
19772         (nds32_long_call_p): Modify.
19773         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19774         * config/nds32/nds32-protos.h
19775         (symbolic_reference_mentioned_p): Declare.
19776         (nds32_legitimize_ict_address): Declare.
19777         (nds32_expand_ict_move): Declare.
19778         (nds32_indirect_call_referenced_p): Declare.
19779         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19780         (nds32_relax_group): Use nds32_ict_const_p as condition.
19781         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19782         (nds32_asm_file_start): Output ict_model directive in asm code.
19783         (nds32_legitimate_address_p): Consider indirect call.
19784         (nds32_print_operand): Consider indirect call.
19785         (nds32_print_operand_address): Consider indirect call.
19786         (nds32_insert_attributes): Handle "indirect_call" attribute.
19787         (TARGET_LEGITIMATE_ADDRESS_P): Define.
19788         (TARGET_LEGITIMATE_CONSTANT_P): Define.
19789         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19790         (TARGET_DELEGITIMIZE_ADDRESS): Define.
19791         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19792         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19793         (TARGET_ICT_MODEL_SMALL): Define.
19794         (TARGET_ICT_MODEL_LARGE): Define.
19795         * config/nds32/nds32.md (movsi): Consider ict model.
19796         (call, call_value): Consider ict model.
19797         (sibcall, sibcall_value): Consider ict model.
19798         * config/nds32/nds32.opt (mict-model): New option.
19799         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19800         model.
19802 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
19803             Monk Chiang  <sh.chiang04@gmail.com>
19804             Jim Wilson <jimw@sifive.com>
19806         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19807         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
19808         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19809         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19810         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19811         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19812         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19813         compute save_libcall_adjustment properly.
19814         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19815         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19816         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19817         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19818         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19819         (ABI_SPEC): Handle mabi=ilp32e.
19820         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19821         (RVE): Add RVE mask.
19822         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19823         <-march>: Add rv32e as an example.
19825 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19827         PR c++/82899
19828         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19829         (intra_create_variable_infos): Handle C++ constructors.
19831 2018-05-18  Martin Liska  <mliska@suse.cz>
19833         * passes.def: Remove a redundant pass.
19835 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19837         PR bootstrap/85838
19838         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19840 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19842         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19843         (ARMv4): Update.
19844         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19845         (ARMv6m): Update.
19846         (armv2, armv2a, armv3, armv3m): Delete architectures.
19847         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19848         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19849         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19850         Delete cpus.
19851         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19852         (*mulsidi3adddi): Likewise.
19853         (mulsidi3): Likewise.
19854         (*mulsidi3_nov6): Likewise.
19855         (umulsidi3): Likewise.
19856         (umulsidi3_nov6): Likewise.
19857         (umaddsidi4): Likewise.
19858         (*umulsidi3adddi): Likewise.
19859         (smulsi3_highpart): Likewise.
19860         (*smulsi3_highpart_nov6): Likewise.
19861         (umulsi3_highpart): Likewise.
19862         (*umulsi3_highpart_nov6): Likewise.
19863         * config/arm/arm.h (arm_arch3m): Delete.
19864         * config/arm/arm.c (arm_arch3m): Delete.
19865         (arm_option_override_internal): Update armv3-related comment.
19866         (arm_configure_build_target): Delete use of isa_bit_mode32.
19867         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19868         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19869         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19870         (mulsa3): Likewise.
19871         (mulusa3): Likewise.
19872         * config/arm/arm-protos.h (arm_arch3m): Delete.
19873         * config/arm/arm-tables.opt: Regenerate.
19874         * config/arm/arm-tune.md: Likewise.
19875         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19876         deleted architectures.
19878 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19880         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19881         (armv5t, armv5te): New features.
19882         (ARMv5, ARMv5e): Delete fgroups.
19883         (ARMv5t, ARMv5te): Adjust for above changes.
19884         (ARMv6m): Likewise.
19885         (armv5, armv5e): Delete arches.
19886         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19887         arm_arch5.
19888         (*call_reg_arm): Likewise.
19889         (*call_value_reg_armv5): Likewise.
19890         (*call_value_reg_arm): Likewise.
19891         (*call_symbol): Likewise.
19892         (*call_value_symbol): Likewise.
19893         (*sibcall_insn): Likewise.
19894         (*sibcall_value_insn): Likewise.
19895         (clzsi2): Likewise.
19896         (prefetch): Likewise.
19897         (define_split and define_peephole2 dependent on arm_arch5):
19898         Likewise.
19899         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19900         arm_arch5e.
19901         (TARGET_ARM_QBIT): Likewise.
19902         (TARGET_DSP_MULTIPLY): Likewise.
19903         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19904         (arm_arch5, arm_arch5e): Delete.
19905         (arm_arch5t, arm_arch5te): Declare.
19906         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19907         (arm_arch5t): Declare.
19908         (arm_option_reconfigure_globals): Update for the above.
19909         (arm_options_perform_arch_sanity_checks): Update comment, replace
19910         use of arm_arch5 with arm_arch5t.
19911         (use_return_insn): Likewise.
19912         (arm_emit_call_insn): Likewise.
19913         (output_return_instruction): Likewise.
19914         (arm_final_prescan_insn): Likewise.
19915         (arm_coproc_builtin_available): Likewise.
19916         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19917         arm_arch5e with arm_arch5t and arm_arch5te.
19918         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19919         (arm_arch5t, arm_arch5te): Declare.
19920         * config/arm/arm-tables.opt: Regenerate.
19921         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19922         * config/arm/t-multilib: Likewise.
19923         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19924         instead of arm_arch5.
19925         (*call_reg_thumb1): Likewise.
19926         (*call_value_reg_thumb1_v5): Likewise.
19927         (*call_value_reg_thumb1): Likewise.
19928         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19929         unreachable path.
19930         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19932 2018-05-18  Martin Liska  <mliska@suse.cz>
19934         PR gcov-profile/84846
19935         * doc/gcov.texi: Document -t option of gcov tool.
19937 2018-05-18  Martin Liska  <mliska@suse.cz>
19939         PR gcov-profile/84846
19940         * gcov.c (print_usage): Add new -t option.
19941         (process_args): Handle the option.
19942         (generate_results): Use stdout as output when requested by
19943         the option.
19945 2018-05-18  Martin Liska  <mliska@suse.cz>
19947         PR gcov-profile/84846
19948         * coverage.c (coverage_init): Write PWD to .gcno file.
19949         * doc/gcov.texi: Document how working directory is printed.
19950         * gcov-dump.c (dump_gcov_file): Print PWD.
19951         * gcov.c (output_intermediate_file): Likewise.
19952         (read_graph_file): Read PWD string.
19953         (output_lines): Print PWD.
19955 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19957         PR middle-end/85817
19958         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19959         for retval and return false if all args to phi are zero.
19961 2018-05-18  Richard Biener  <rguenther@suse.de>
19963         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19964         method.
19965         (evrp_dom_walker::before_dom_children): Call it.
19967 2018-05-18  Richard Biener  <rguenther@suse.de>
19969         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19970         results when processing array refs with variable index.
19972 2018-05-18  Toon Moene  <toon@moene.org>
19974         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19975         directly after that of -floop-interchange. Indicate that both
19976         options are enabled by default when specifying -O3.
19978 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19980         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19981         iterator.  Delete separate integer-mode vec_set<mode> expander.
19982         (aarch64_simd_vec_setv2di): Delete.
19983         (vec_setv2di): Delete.
19984         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19985         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19986         the "w, r" alternative.
19988 2018-05-18  Martin Liska  <mliska@suse.cz>
19990         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19991         * tree-pass.h (make_pass_lower_switch_O0): New function.
19992         * tree-switch-conversion.c (node_has_low_bound): Remove.
19993         (node_has_high_bound): Likewise.
19994         (node_is_bounded): Likewise.
19995         (class pass_lower_switch): Make it a template type and create
19996         two instances.
19997         (pass_lower_switch::execute): Add template argument.
19998         (make_pass_lower_switch): New function.
19999         (make_pass_lower_switch_O0): New function.
20000         (do_jump_if_equal): Remove.
20001         (emit_case_nodes): Simplify to just handle all 3 cases and leave
20002         all the hard work to tree optimization passes.
20004 2018-05-18  Martin Liska  <mliska@suse.cz>
20006         * dbgcnt.c (limit_low): Renamed from limit.
20007         (limit_high): New variable.
20008         (dbg_cnt_is_enabled): Check for upper limit.
20009         (dbg_cnt): Adjust dumping.
20010         (dbg_cnt_set_limit_by_index): Add new argument for high
20011         value.
20012         (dbg_cnt_set_limit_by_name): Likewise.
20013         (dbg_cnt_process_single_pair): Parse new format.
20014         (dbg_cnt_process_opt): Use strtok.
20015         (dbg_cnt_list_all_counters): Remove 'value' and add
20016         'limit_high'.
20017         * doc/invoke.texi: Document changes.
20019 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
20021         * doc/sourcebuild.texi (scalar_all_fma): Document.
20022         * tree.def (FMA_EXPR): Delete.
20023         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
20024         * internal-fn.c (ternary_direct): New macro.
20025         (expand_ternary_optab_fn): Likewise.
20026         (direct_ternary_optab_supported_p): Likewise.
20027         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
20028         * builtins.c (fold_builtin_fma): Delete.
20029         (fold_builtin_3): Don't call it.
20030         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
20031         * expr.c (expand_expr_real_2): Likewise.
20032         * fold-const.c (operand_equal_p): Likewise.
20033         (fold_ternary_loc): Likewise.
20034         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
20035         * gimple.c (DEFTREECODE): Likewise.
20036         * gimplify.c (gimplify_expr): Likewise.
20037         * optabs-tree.c (optab_for_tree_code): Likewise.
20038         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20039         * tree-eh.c (operation_could_trap_p): Likewise.
20040         (stmt_could_throw_1_p): Likewise.
20041         * tree-inline.c (estimate_operator_cost): Likewise.
20042         * tree-pretty-print.c (dump_generic_node): Likewise.
20043         (op_code_prio): Likewise.
20044         * tree-ssa-loop-im.c (stmt_cost): Likewise.
20045         * tree-ssa-operands.c (get_expr_operands): Likewise.
20046         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
20047         * fold-const-call.h (fold_fma): Delete.
20048         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
20049         CFN_FNMA and CFN_FNMS.
20050         (fold_fma): Delete.
20051         * genmatch.c (combined_fn): New enum.
20052         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
20053         (commutative_op): New function.
20054         (commutate): Use it.  Handle more than 2 operands.
20055         (dt_operand::gen_gimple_expr): Use commutative_op.
20056         (parser::parse_expr): Allow :c to be used with non-binary
20057         operators if the commutative operand is known.
20058         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
20059         CFN_FMS, CFN_FNMA and CFN_FNMS.
20060         (backprop::process_assign_use): Remove FMA_EXPR handling.
20061         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
20062         (gen_hsa_fma): New function.
20063         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
20064         IFN_FNMA and IFN_FNMS.
20065         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
20066         * gimple-fold.h (follow_all_ssa_edges): Declare.
20067         * gimple-fold.c (follow_all_ssa_edges): New function.
20068         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
20069         gimple_build interface and use follow_all_ssa_edges to fold the result.
20070         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
20071         instead of checking for optabs directly.
20072         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
20073         rather than FMA_EXPRs.
20074         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
20075         call to IFN_FMA instead of an FMA_EXPR.
20077 2018-05-17  Jim Wilson  <jimw@sifive.com>
20079         * expr.c (do_tablejump): When converting index to Pmode, if we have a
20080         sign extended promoted subreg, and the range does not have the sign bit
20081         set, then do a sign extend.
20083         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
20084         test, check for sign extended subreg and/or constant operands, and
20085         do a sign extend in that case.
20087 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
20089         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
20090         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
20091         Add untyped.
20092         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
20093         Change logics_shift_reg to logics_shift_imm.
20094         (thunderx2t99_fp_loadpair_basic): Delete.
20095         (thunderx2t99_fp_storepair_basic): Delete.
20096         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
20097         (thunderx2t99_asimd_polynomial): Delete.
20098         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
20099         and neon_fp_mul_d_scalar_q.
20100         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
20101         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
20102         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
20103         (thunderx2t99_asimd_lut): Add missing tbl types.
20104         (thunderx2t99_asimd_ext): Delete.
20105         (thunderx2t99_asimd_load1_1_mult): Delete.
20106         (thunderx2t99_asimd_load1_2_mult): Delete.
20107         (thunderx2t99_asimd_load1_ldp): New.
20108         (thunderx2t99_asimd_load1): New.
20109         (thunderx2t99_asimd_load2): Add missing *load2* types.
20110         (thunderx2t99_asimd_load3): New.
20111         (thunderx2t99_asimd_load4): New.
20112         (thunderx2t99_asimd_store1_1_mult): Delete.
20113         (thunderx2t99_asimd_store1_2_mult): Delete.
20114         (thunderx2t99_asimd_store2_mult): Delete.
20115         (thunderx2t99_asimd_store2_onelane): Delete.
20116         (thunderx2t99_asimd_store_stp): New.
20117         (thunderx2t99_asimd_store1): New.
20118         (thunderx2t99_asimd_store2): New.
20119         (thunderx2t99_asimd_store3): New.
20120         (thunderx2t99_asimd_store4): New.
20122 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
20124         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
20125         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
20127 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
20128             Segher Boessenkool  <segher@kernel.crashing.org>
20130         PR target/85698
20131         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
20132         operand.
20134 2018-05-17  Richard Biener  <rguenther@suse.de>
20136         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
20137         for pruning loop and prune defs feeding only already visited PHIs.
20139 2018-05-17  Richard Biener  <rguenther@suse.de>
20141         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
20143 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
20144             Richard Biener  <rguenther@suse.de>
20146         PR tree-optimization/85793
20147         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
20148         for VMAT_ELEMENTWISE.
20150 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
20152         * internal-fn.h (lookup_internal_fn): Declare
20153         * internal-fn.c (lookup_internal_fn): New function.
20154         * gimple.c (gimple_build_call_from_tree): Handle calls to
20155         internal functions.
20156         * gimple-pretty-print.c (dump_gimple_call): Print "." before
20157         internal function names.
20158         * tree-pretty-print.c (dump_generic_node): Likewise.
20159         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
20161 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
20163         * gimple-fold.h (gimple_build): Make the function forms take
20164         combined_fn rather than built_in_function.
20165         (gimple_simplify): Likewise.
20166         * gimple-match-head.c (gimple_simplify): Likewise.
20167         * gimple-fold.c (gimple_build): Likewise.
20168         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
20169         rather than gimple_build_call_internal.
20170         (get_initial_defs_for_reduction): Likewise.
20171         (vect_create_epilog_for_reduction): Likewise.
20172         (vectorizable_live_operation): Likewise.
20174 2018-05-17  Martin Liska  <mliska@suse.cz>
20176         * gimple-ssa-sprintf.c (format_directive): Do not use
20177         space in between 'G_' and '('.
20179 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
20181         PR target/85323
20182         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
20183         even if the mask is not all ones.
20185         PR target/85323
20186         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
20187         vector.
20188         (ix86_gimple_fold_builtin): Likewise.
20190         PR target/85323
20191         * config/i386/i386.c: Include tree-vector-builder.h.
20192         (ix86_vector_shift_count): New function.
20193         (ix86_fold_builtin): Fold shift builtins by scalar count.
20194         (ix86_gimple_fold_builtin): Likewise.
20196         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
20197         _mm512_setzero): New intrinsics.
20199 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
20200             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20202         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
20203         code generation for cases where splatting a value is not useful.
20204         * simplify-rtx.c (simplify_ternary_operation): Simplify
20205         vec_merge across a vec_duplicate and a paradoxical subreg forming
20206         a vector mode to a vec_concat.
20208 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
20210         * config.gcc: Support "goldmont-plus".
20211         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20212         "goldmont-plus".
20213         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20214         PROCESSOR_GOLDMONT_PLUS.
20215         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
20216         (processor_target_table): Add "goldmont-plus".
20217         (PTA_GOLDMONT_PLUS): Define.
20218         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
20219         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
20220         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
20221         (fold_builtin_cpu): Add "goldmont-plus".
20222         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
20223         (ix86_option_override_internal): Add "goldmont-plus".
20224         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
20225         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
20226         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
20227         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
20229 2018-05-17  Richard Biener  <rguenther@suse.de>
20231         PR tree-optimization/85757
20232         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
20233         remove defs that only feed that PHI from further processing.
20235 2018-05-16  Jim Wilson  <jimw@sifive.com>
20237         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
20238         asterisk to name.
20239         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
20241 2018-05-16  Mark Wielaard  <mark@klomp.org>
20243         * dwarf2out.c (count_index_strings): New function.
20244         (output_indirect_strings): Call count_index_strings and generate
20245         header for dwarf_version >= 5.
20247 2018-05-16  Mark Wielaard  <mark@klomp.org>
20249         * dwarf2out.c (dwarf_FORM): New function.
20250         (set_indirect_string): Use dwarf_FORM.
20251         (reset_indirect_string): Likewise.
20252         (size_of_die): Likewise.
20253         (value_format): Likewise.
20254         (output_die): Likewise.
20255         (add_skeleton_AT_string): Likewise.
20256         (output_macinfo_op): Likewise.
20257         (index_string): Likewise.
20258         (output_index_string_offset): Likewise.
20259         (output_index_string): Likewise.
20260         (count_index_strings): Likewise.
20262 2018-05-16  Carl Love  <cel@us.ibm.com>
20264         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20265         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
20267 2018-05-16  Martin Jambor  <mjambor@suse.cz>
20269         * ipa-prop.c (ipa_free_all_edge_args): Remove.
20270         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
20272 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
20274         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
20275         (fnma<mode>4): Likewise.
20276         (fms<mode>4): Likewise.
20277         (fnms<mode>4): Likewise.
20278         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
20279         (aarch64_fnma<mode>4): Likewise.
20280         (aarch64_fms<mode>4): Likewise.
20281         (aarch64_fnms<mode>4): Likewise.
20282         (aarch64_fnmadd<mode>4): Likewise.
20284 2018-05-16  Jason Merrill  <jason@redhat.com>
20286         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
20288 2018-05-16  Richard Biener  <rguenther@suse.de>
20290         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
20291         (dump_stmt_cost): Declare.
20292         (add_stmt_cost): Dump cost we add.
20293         (add_stmt_costs): New function.
20294         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
20295         No longer exported.
20296         (vect_analyze_stmt): Adjust prototype.
20297         (vectorizable_condition): Likewise.
20298         (vectorizable_live_operation): Likewise.
20299         (vectorizable_reduction): Likewise.
20300         (vectorizable_induction): Likewise.
20301         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
20302         cost vector to pass to vectorizable_ and record afterwards.
20303         (vect_model_reduction_cost): Take cost vector argument and adjust.
20304         (vect_model_induction_cost): Likewise.
20305         (vectorizable_reduction): Likewise.
20306         (vectorizable_induction): Likewise.
20307         (vectorizable_live_operation): Likewise.
20308         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
20309         SLP_TREE_NUMBER_OF_VEC_STMTS.
20310         (vect_analyze_slp_cost_1): Remove.
20311         (vect_analyze_slp_cost): Likewise.
20312         (vect_slp_analyze_node_operations): Take visited args and
20313         a target cost vector.  Avoid processing already visited stmt sets.
20314         (vect_slp_analyze_operations): Use a local cost vector to gather
20315         costs and register those of non-discarded instances.
20316         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
20317         (vect_schedule_slp_instance): Remove copying of
20318         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
20319         zero.
20320         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
20321         adding cost.  Record cost entry location.
20322         (vect_prologue_cost_for_slp_op): Function to compute cost of
20323         a constant or invariant generated for SLP vect in the prologue,
20324         split out from vect_analyze_slp_cost_1.
20325         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
20326         (vect_model_promotion_demotion_cost): Likewise.
20327         (vect_model_store_cost): Likewise, make static.
20328         (vect_model_load_cost): Likewise.
20329         (vectorizable_bswap): Add cost vector arg and adjust.
20330         (vectorizable_call): Likewise.
20331         (vectorizable_simd_clone_call): Likewise.
20332         (vectorizable_conversion): Likewise.
20333         (vectorizable_assignment): Likewise.
20334         (vectorizable_shift): Likewise.
20335         (vectorizable_operation): Likewise.
20336         (vectorizable_store): Likewise.
20337         (vectorizable_load): Likewise.
20338         (vectorizable_condition): Likewise.
20339         (vectorizable_comparison): Likewise.
20340         (can_vectorize_live_stmts): Likewise.
20341         (vect_analyze_stmt): Likewise.
20342         (vect_transform_stmt): Adjust calls to vectorizable_*.
20343         * tree-vectorizer.c: Include gimple-pretty-print.h.
20344         (dump_stmt_cost): New function.
20346 2018-05-16  Richard Biener  <rguenther@suse.de>
20348         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
20349         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
20350         * tree-ssa-dse.c: Include tree-ssa-loop.h.
20351         (check_name): New callback.
20352         (dse_classify_store): Track cycles via a visited bitmap of PHI
20353         defs and simplify handling of in-loop and across loop dead stores
20354         and properly fail for loop-variant refs.  Handle byte-tracking with
20355         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
20356         limiting the walk.
20358 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
20360         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
20361         (vect_get_mask_type_for_stmt): Likewise.
20362         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
20363         split out from...
20364         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
20365         to determine the statement's vector type and the vector type that
20366         should be used for calculating nunits.  Deal with cases in which
20367         the type has to be deferred.
20368         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
20369         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
20370         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
20371         (vect_determine_vf_for_stmt): New functions, split out from...
20372         (vect_determine_vectorization_factor): ...here.
20373         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
20374         (vect_get_mask_type_for_stmt): New functions, split out from
20375         vect_determine_vectorization_factor.
20377 2018-05-16  Richard Biener  <rguenther@suse.de>
20379         * tree-cfg.c (verify_gimple_assign_ternary): Properly
20380         verify the [VEC_]COND_EXPR embedded comparison.
20382 2018-05-15  Martin Sebor  <msebor@redhat.com>
20384         PR tree-optimization/85753
20385         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
20386         RECORD_TYPE in addition to ARRAY_TYPE.
20388 2018-05-15  Martin Sebor  <msebor@redhat.com>
20390         PR middle-end/85643
20391         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
20393 2018-05-15  Richard Biener  <rguenther@suse.de>
20395         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
20396         add by_clobber_p one.  Change algorithm to collect all defs
20397         representing uses we need to walk and try reducing them to
20398         a single one before failing.
20399         (dse_dom_walker::dse_optimize_stmt): Adjust.
20401 2018-05-13  Mark Wielaard  <mark@klomp.org>
20403         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
20404         (size_of_loc_descr): Likewise.
20405         (output_loc_operands): Likewise.
20406         (output_loc_operands_raw): Likewise.
20407         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
20408         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
20409         (hash_loc_operands): Likewise.
20410         (compare_loc_operands): Likewise.
20412 2018-05-14  Mark Wielaard  <mark@klomp.org>
20414         * dwarf2out.c (count_index_addrs): New function.
20415         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
20417 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20419         PR tree-optimization/83648
20420         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
20421         return value as malloc candidate.
20423 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20425         PR ipa/85734
20426         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
20427         param as true in call to suggest_attribute.
20429 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
20431         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
20432         -mreadonly-in-sdata.
20434 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20436         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
20437         New pattern.
20438         (aarch64_crypto_aesd_fused): Likewise.
20440 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
20442         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
20443         (movsi_aarch64): Likewise.
20444         (load_pairsi): Likewise.
20445         (load_pairdi): Likewise.
20446         (store_pairsi): Likewise.
20447         (store_pairdi): Likewise.
20448         (load_pairsf): Likewise.
20449         (load_pairdf): Likewise.
20450         (store_pairsf): Likewise.
20451         (store_pairdf): Likewise.
20452         (zero_extend): Likewise.
20453         (trunc): Swap alternatives.
20454         (fcvt_target): Add '?' to prefer w over r.
20456 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
20458         PR target/85756
20459         * config/i386/i386.md: Disallow non-commutative arithmetics in
20460         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
20461         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
20462         in the peephole2 before it.
20464 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
20466         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
20467         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
20468         (ix86_handle_option): Handle -mcldemote.
20469         * config.gcc: New header.
20470         * config/i386/cldemoteintrin.h: New file.
20471         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
20472         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20473         -mcldemote.
20474         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20475         OPTION_MASK_ISA_CLDEMOTE.
20476         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
20477         (ix86_valid_target_attribute_inner_p): Ditto.
20478         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
20479         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
20480         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
20481         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
20482         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
20483         (cldemote): New.
20484         * config/i386/i386.opt: Add -mcldemote.
20485         * config/i386/x86intrin.h: New header.
20486         * doc/invoke.texi: Add -mcldemote.
20488 2018-05-14  Richard Biener  <rguenther@suse.de>
20490         * doc/match-and-simplify.texi: Adjust :s documentation.
20492 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
20494         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
20495         intended memcpy size.
20496         (REORDER_45): Likewise.
20498 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
20500         * sort.cc: New file.
20501         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
20502         * vec.c (qsort_chk): Use gcc_qsort.
20503         * Makefile.in (OBJS-libcommon): Add sort.o.
20504         (build/sort.o): New target.  Use it...
20505         (BUILD_RTL): ... here, and...
20506         (build/gencfn-macros): ... here, and...
20507         (build/genmatch): ... here.
20509 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
20510             Chung-Ju Wu  <jasonwucj@gmail.com>
20512         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
20513         * config/nds32/nds32-graywolf.md: New file.
20514         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
20515         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
20516         pipeline.
20517         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
20518         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
20519         * config/nds32/nds32.md (pipeline_model): Add graywolf.
20520         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
20521         * config/nds32/pipelines.md: Include n15 settings.
20523 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
20524             Chung-Ju Wu  <jasonwucj@gmail.com>
20526         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
20527         * config/nds32/nds32-n13.md: New file.
20528         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
20529         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
20530         pipeline.
20531         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
20532         * config/nds32/nds32.md (pipeline_model): Add n13.
20533         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
20534         * config/nds32/pipelines.md: Include n13 settings.
20536 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
20537             Chung-Ju Wu  <jasonwucj@gmail.com>
20539         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
20540         * config/nds32/nds32-n10.md: New file.
20541         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
20542         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
20543         pipeline.
20544         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
20545         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
20546         * config/nds32/nds32.md (pipeline_model): Add n10.
20547         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
20548         * config/nds32/pipelines.md: Include n10 settings.
20550 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
20551             Kito Cheng  <kito.cheng@gmail.com>
20552             Chung-Ju Wu  <jasonwucj@gmail.com>
20554         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
20555         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20556         Add enum values for DSP extension instructions.
20557         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
20558         New constraints.
20559         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
20560         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
20561         New code iterators.
20562         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
20563         * config/nds32/nds32-dspext.md: New file for DSP implementation.
20564         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
20565         * config/nds32/nds32-intrinsic.md: Likewise.
20566         * config/nds32/nds32_intrinsic.h: Likewise.
20567         * config/nds32/nds32-md-auxiliary.c: Likewise.
20568         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
20569         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
20570         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
20571         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
20572         * config/nds32/nds32-protos.h: New declarations for DSP extension.
20573         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
20574         TYPE_DMAC in switch statement.
20575         * config/nds32/nds32.c: New checking and implementation for DSP
20576         extension instructions.
20577         * config/nds32/nds32.h: Likewise.
20578         * config/nds32/nds32.md: Likewise.
20579         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
20580         * config/nds32/predicates.md: Implement new predicates for DSP
20581         extension.
20583 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
20585         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
20586         Reformat alternatives and attributes so it is easier to identify
20587         which constraints/attributes go with which instruction.
20588         (mov<mode>_hardfloat32, FMOVE64): Likewise.
20589         (mov<mode>_softfloat32, FMOVE64): Likewise.
20590         (mov<mode>_hardfloat64, FMOVE64): Likewise.
20591         (mov<mode>_softfloat64, FMOVE64): Likewise.
20593 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20595         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20596         subsection.
20597         (Basic PowerPC Built-in Functions): The new name of the
20598         subsection previously known as "PowerPC Built-in Functions".
20599         (Basic PowerPC Built-in Functions Available on all Configurations):
20600         New subsubsection.
20601         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
20602         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
20603         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
20604         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
20606 2018-05-11  Martin Jambor  <mjambor@suse.cz>
20608         PR ipa/85655
20609         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
20610         single const.
20612 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
20614         PR target/85733
20615         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
20617 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
20619         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
20620         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
20621         (ix86_handle_option): Handle -mwaitpkg.
20622         * config.gcc: New header.
20623         * config/i386/cpuid.h (bit_WAITPKG): New bit.
20624         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
20625         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
20626         function type.
20627         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20628         OPTION_MASK_ISA_WAITPKG.
20629         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
20630         (ix86_option_override_internal): Add PTA_WAITPKG.
20631         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
20632         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
20633         IX86_BUILTIN_TPAUSE.
20634         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
20635         __builtin_ia32_umwait and __builtin_ia32_tpause.
20636         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
20637         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
20638         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
20639         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
20640         UNSPECV_TPAUSE): New.
20641         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
20642         * config/i386/i386.opt: Add -mwaitpkg.
20643         * config/i386/waitpkgintrin.h: New file.
20644         * config/i386/x86intrin.h: New header.
20645         * doc/invoke.texi: Add -mwaitpkg.
20647 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
20649         PR target/85606
20650         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
20651         equivalent.
20652         (cortex-m0): Use armv6s-m isa.
20653         (cortex-m0plus): Likewise.
20654         (cortex-m1): Likewise.
20655         (cortex-m0.small-multiply): Likewise.
20656         (cortex-m0plus.small-multiply): Likewise.
20657         (cortex-m1.small-multiply): Likewise.
20659 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
20660             Jakub Jelinek  <jakub@redhat.com>
20662         PR tree-optimization/85692
20663         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
20664         source permute as well.
20666 2018-05-11  Martin Liska  <mliska@suse.cz>
20668         PR sanitizer/85556
20669         * doc/extend.texi: Document LLVM style format for no_sanitize
20670         attribute.
20672 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
20674         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
20675         mode_supports_vsx_dform_quad to mode_supports_dq_form.
20676         (mode_supports_vsx_dform_quad): Likewise.
20677         (mode_supports_vmx_dform): Move these functions to be next to the
20678         other mode_supports functions.
20679         (mode_supports_dq_form): Likewise.
20680         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
20681         mode_supports_dq_form.
20682         (reg_offset_addressing_ok_p): Likewise.
20683         (offsettable_ok_by_alignment): Likewise.
20684         (rs6000_legitimate_offset_address_p): Likewise.
20685         (legitimate_lo_sum_address_p): Likewise.
20686         (rs6000_legitimize_address): Likewise.
20687         (rs6000_legitimize_reload_address): Likewise.
20688         (rs6000_secondary_reload_inner): Likewise.
20689         (rs6000_preferred_reload_class): Likewise.
20690         (rs6000_output_move_128bit): Likewise.
20692 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20694         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20695         Generate SImode target register for null target.
20696         <case IX86_BUILTIN_XGETBV>: Ditto.
20697         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20698         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20700 2018-05-10  Carl Love  <cel@us.ibm.com>
20702         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20703         dcbtt and dcbtstt if operands[2] is 0.
20705 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20707         PR target/85693
20708         * config/i386/sse.md (usadv64qi): New expander.
20710 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
20712         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20713         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20714         -maltivec=be support.
20715         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20716         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20717         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20718         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20719         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20720         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20721         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20722         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20723         altivec_vsumsws): Adjust.
20724         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20725         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20726         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20727         support.
20728         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20729         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20730         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20731         (altivec_lve<VI_char>x): Delete expand.
20732         (*altivec_lve<VI_char>x_internal): Rename to...
20733         (altivec_lve<VI_char>x): ... this.
20734         (altivec_lvxl_<mode>): Delete expand.
20735         (*altivec_lvxl_<mode>_internal): Rename to ...
20736         (altivec_lvxl_<mode>): ... this.
20737         (altivec_stvxl_<mode>): Delete expand.
20738         (*altivec_stvxl_<mode>_internal): Rename to ...
20739         (altivec_stvxl_<mode>): ... this.
20740         (altivec_stve<VI_char>x): Delete expand.
20741         (*altivec_stve<VI_char>x_internal): Rename to ...
20742         (altivec_stve<VI_char>x): ... this.
20743         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20744         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20745         reduc_plus_scal_<mode>): Adjust.
20746         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20747         comment.
20748         (rs6000_cpu_cpp_builtins): Adjust.
20749         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20750         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20751         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20752         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20753         -maltivec=be support.
20754         (rs6000_split_vec_extract_var): Adjust.
20755         (rs6000_split_v4si_init): Adjust.
20756         (swap_selector_for_mode): Delete.
20757         (altivec_expand_lvx_be, altivec_expand_stvx_be,
20758         altivec_expand_stvex_be): Delete.
20759         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20760         -maltivec=be support.
20761         (rs6000_gimple_fold_builtin): Ditto.
20762         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20763         Adjust.
20764         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20765         (TARGET_DIRECT_MOVE_64BIT): Adjust.
20766         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20767         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20768         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20769         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20770         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20771         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20772         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20773         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20774         anonymous split): Adjust.
20775         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20776         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20778 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
20780         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20781         when --with-gxx-include-dir is also specified.
20782         * configure: Regenerate.
20784 2018-05-09  Jim Wilson  <jimw@sifive.com>
20786         PR target/84797
20787         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20788         * config/riscv/t-withmultilib: New.
20789         * config/riscv/withmultilib.h: New.
20790         * doc/install.texi: Document RISC-V --with-multilib-list support.
20792 2018-05-09  Richard Biener  <rguenther@suse.de>
20794         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20795         vector.
20796         (vect_bb_vectorization_profitable_p): Adjust.  Compute
20797         actual scalar cost using the cost vector and the add_stmt_cost
20798         machinery.
20800 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20802         PR rtl-optimization/85645
20803         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20804         in the REG_CFA_REGISTER note for LR, don't leave it empty.
20806 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20808         PR rtl-optimization/85645
20809         * shrink-wrap.c (spread_components): Return a boolean saying if
20810         anything was changed.
20811         (try_shrink_wrapping_separate): Iterate spread_components until
20812         nothing changes anymore.
20814 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20816         PR rtl-optimization/85645
20817         * regrename.c (build_def_use): Also kill the chains that include the
20818         destination of a REG_CFA_REGISTER note.
20820 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20822         PR rtl-optimization/85645
20823         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20824         insn that has a REG_CFA_REGISTER note.
20826 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20828         * cfgexpand.c (expand_clobber): New function.
20829         (expand_gimple_stmt_1): Use it.
20830         * tree-vect-stmts.c (vect_clobber_variable): New function,
20831         split out from...
20832         (vectorizable_simd_clone_call): ...here.
20833         (vectorizable_store): Emit a clobber either side of an
20834         IFN_STORE_LANES sequence.
20835         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20837 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20839         PR target/85626
20840         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20841         (define_insn "trap_if_false"): Add exit after trap.
20843 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20845         PR rtl-optimization/85638
20846         * bb-reorder.c: Include common/common-target.h.
20847         (create_forwarder_block): New function extracted from...
20848         (fix_up_crossing_landing_pad): ...here.  Rename into...
20849         (dw2_fix_up_crossing_landing_pad): ...this.
20850         (sjlj_fix_up_crossing_landing_pad): New function.
20851         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20852         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20853         from both partitions and exit the loop after one iteration.
20855 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20857         Revert:
20858         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20859         subsection.
20860         (Basic PowerPC Built-in Functions): The new name of the
20861         subsection previously known as "PowerPC Built-in Functions".
20862         (Basic PowerPC Built-in Functions Available on all Configurations):
20863         New subsubsection.
20864         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20865         subsubsection.
20866         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20867         subsubsection.
20868         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20869         subsubsection.
20870         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20871         subsubsection.
20873 2018-05-08  Jim Wilson  <jimw@sifive.com>
20875         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20876         (LD_EMUL_SUFFIX): New.
20877         (LINK_SPEC): Use it.
20879 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20881         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20882         subsection.
20883         (Basic PowerPC Built-in Functions): The new name of the
20884         subsection previously known as "PowerPC Built-in Functions".
20885         (Basic PowerPC Built-in Functions Available on all Configurations):
20886         New subsubsection.
20887         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20888         subsubsection.
20889         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20890         subsubsection.
20891         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20892         subsubsection.
20893         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20894         subsubsection.
20896 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20898         PR target/85683
20899         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20900         after cmpelim optimization.
20902 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20904         * config.gcc: Support "goldmont".
20905         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20906         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20907         PROCESSOR_GOLDMONT.
20908         * config/i386/i386.c (m_GOLDMONT): Define.
20909         (processor_target_table): Add "goldmont".
20910         (PTA_GOLDMONT): Define.
20911         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20912         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20913         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20914         (fold_builtin_cpu): Add "goldmont".
20915         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20916         (ix86_option_override_internal): Add "goldmont".
20917         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20918         (processor_type): Add PROCESSOR_GOLDMONT.
20919         * config/i386/i386.md: Add CPU "glm".
20920         * config/i386/glm.md: New file.
20921         * config/i386/x86-tune.def: Add m_GOLDMONT.
20922         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20924 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20926         PR target/85572
20927         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20928         E_V4DImode.
20929         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20930         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20931         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20933         PR target/85317
20934         * config/i386/i386.c (ix86_fold_builtin): Handle
20935         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20937         PR target/85480
20938         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20939         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20941 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20943         PR target/85658
20944         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20945         (check_arch): Likewise.
20946         (check_fpu): Return the result rather than printing it.
20947         (end arch): Fix operator precedence.
20948         (end cpu): Likewise.
20949         (END): Print the result from check_fpu.
20951 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20952             Alan Hayward  <alan.hayward@arm.com>
20953             David Sherwood  <david.sherwood@arm.com>
20955         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20956         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20957         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20958         (*fcmuo<mode>_and): New patterns.
20960 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20962         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20963         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20964         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20965         (cmp_op, sve_imm_con): New code attributes.
20966         (SVE_COND_INT_CMP, imm_con): Delete.
20967         (cmp_op): Remove above unspecs from int attribute.
20968         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20969         to...
20970         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20971         comparison-specific unspecs.
20972         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20973         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20974         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20975         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20976         (*vec_fcm<cmp_op><mode>): Rename to...
20977         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20978         (*vec_fcmuo<mode>): Rename to...
20979         (*fcmuo<mode>): ...this and adjust likewise.
20980         (*pred_fcm<cmp_op><mode>): New pattern.
20981         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20982         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20983         functions.
20984         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20985         and UNORDERED.
20986         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20987         (aarch64_emit_sve_predicated_cond): New function.
20988         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20989         (aarch64_emit_unspec_cond_or): Replace with...
20990         (aarch64_emit_sve_or_conds): ...this new function.  Use
20991         aarch64_emit_sve_ptrue_op for the individual comparisons and
20992         aarch64_emit_binop to OR them together.
20993         (aarch64_emit_inverted_unspec_cond): Replace with...
20994         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20995         aarch64_emit_sve_ptrue_op for the comparison and
20996         aarch64_emit_unop to invert the result.
20997         (aarch64_expand_sve_vec_cmp_float): Update after the above
20998         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
21000 2018-05-07  Nathan Sidwell  <nathan@acm.org>
21002         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
21003         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
21004         (Backwards Compatibility): Likewise.
21006 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21008         PR bootstrap/85681
21009         Revert:
21010         2018-05-07  Luis Machado  <luis.machado@linaro.org>
21012         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21013         <prefetch_dynamic_strides>: New const bool field.
21014         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21015         prefetch_dynamic_strides.
21016         (exynosm1_prefetch_tune): Likewise.
21017         (thunderxt88_prefetch_tune): Likewise.
21018         (thunderx_prefetch_tune): Likewise.
21019         (thunderx2t99_prefetch_tune): Likewise.
21020         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21021         to false.
21022         (aarch64_override_options_internal): Update to set
21023         PARAM_PREFETCH_DYNAMIC_STRIDES.
21024         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21025         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21026         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21027         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21028         prefetch-dynamic-strides setting.
21030         2018-05-07  Luis Machado  <luis.machado@linaro.org>
21032         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21033         <minimum_stride>: New const int field.
21034         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21035         minimum_stride field.
21036         (exynosm1_prefetch_tune): Likewise.
21037         (thunderxt88_prefetch_tune): Likewise.
21038         (thunderx_prefetch_tune): Likewise.
21039         (thunderx2t99_prefetch_tune): Likewise.
21040         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21041         (aarch64_override_options_internal): Update to set
21042         PARAM_PREFETCH_MINIMUM_STRIDE.
21043         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21044         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21045         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21046         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21047         stride is constant and is below the minimum stride threshold.
21049 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21051         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
21052         to 512.
21054 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21056         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21057         <prefetch_dynamic_strides>: New const bool field.
21058         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21059         prefetch_dynamic_strides.
21060         (exynosm1_prefetch_tune): Likewise.
21061         (thunderxt88_prefetch_tune): Likewise.
21062         (thunderx_prefetch_tune): Likewise.
21063         (thunderx2t99_prefetch_tune): Likewise.
21064         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21065         to false.
21066         (aarch64_override_options_internal): Update to set
21067         PARAM_PREFETCH_DYNAMIC_STRIDES.
21068         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21069         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21070         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21071         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21072         prefetch-dynamic-strides setting.
21074 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21076         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21077         <minimum_stride>: New const int field.
21078         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21079         minimum_stride field.
21080         (exynosm1_prefetch_tune): Likewise.
21081         (thunderxt88_prefetch_tune): Likewise.
21082         (thunderx_prefetch_tune): Likewise.
21083         (thunderx2t99_prefetch_tune): Likewise.
21084         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21085         (aarch64_override_options_internal): Update to set
21086         PARAM_PREFETCH_MINIMUM_STRIDE.
21087         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21088         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21089         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21090         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21091         stride is constant and is below the minimum stride threshold.
21093 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
21095         PR c++/85659
21096         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
21097         the type is addressable.  Don't force op into register if it has
21098         BLKmode.
21100 2018-05-05  Roland McGrath  <mcgrathr@google.com>
21102         PR other/77609
21103         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
21104         any section for which we don't know a specific type it should have,
21105         regardless of name.  Previously this was done only for the exact
21106         names ".init_array", ".fini_array", and ".preinit_array".
21107         (default_elf_asm_named_section): Add comment about
21108         relationship with default_section_type_flags and SECTION_NOTYPE.
21109         (get_section): Don't consider it a type conflict if one side has
21110         SECTION_NOTYPE and the other doesn't, as long as neither has the
21111         SECTION_BSS et al used in the default_section_type_flags logic.
21113 2018-05-05  Tom de Vries  <tom@codesourcery.com>
21115         PR target/85653
21116         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
21117         (workaround_barsyncs): New function.
21118         (nvptx_reorg): Use workaround_barsyncs.
21119         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21120         (define_expand "nvptx_membar_cta"): New define_expand.
21121         (define_insn "*nvptx_membar_cta"): New insn.
21123 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
21125         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
21126         To improve optimization opportunities.
21127         * builtin-types.def: The new needed builtin types for the above.
21129 2018-05-04  Richard Biener  <rguenther@suse.de>
21131         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
21132         * gimple-ssa-store-merging.c
21133         (imm_store_chain_info::output_merged_store): Remove redundant create,
21134         release split_store vector contents on failure.
21135         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
21136         scalar stmt vector on cache hit.
21138 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
21140         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
21141         Xilinx FP support.
21142         * config.gcc (powerpc-xilinx-eabi*): Remove.
21143         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
21144         support.
21145         (fusion_addis_mem_combo_load): Ditto.
21146         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
21147         FP support.
21148         (rs6000_cpu_cpp_builtins): Ditto.
21149         * config/rs6000/rs6000-linux.c
21150         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
21151         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
21152         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
21153         support.
21154         (rs6000_setup_reg_addr_masks): Ditto.
21155         (rs6000_init_hard_regno_mode_ok): Ditto.
21156         (rs6000_option_override_internal): Ditto.
21157         (legitimate_lo_sum_address_p): Ditto.
21158         (rs6000_legitimize_address): Ditto.
21159         (rs6000_legitimize_reload_address): Ditto.
21160         (rs6000_legitimate_address_p): Ditto.
21161         (abi_v4_pass_in_fpr): Ditto.
21162         (setup_incoming_varargs): Ditto.
21163         (rs6000_gimplify_va_arg): Ditto.
21164         (rs6000_split_multireg_move): Ditto.
21165         (rs6000_savres_strategy): Ditto.
21166         (rs6000_emit_prologue_components): Ditto.
21167         (rs6000_emit_epilogue_components): Ditto.
21168         (rs6000_emit_prologue): Ditto.
21169         (rs6000_emit_epilogue): Ditto.
21170         (rs6000_elf_file_end): Ditto.
21171         (rs6000_function_value): Ditto.
21172         (rs6000_libcall_value): Ditto.
21173         * config/rs6000/rs6000.h: Ditto.
21174         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
21175         (TARGET_MINMAX): ... this.  New.
21176         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
21177         * config/rs6000/rs6000.md: Remove Xilinx FP support.
21178         (*movsi_internal1_single): Delete.
21179         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
21180         mfpu=, mxilinx-fpu): Delete.
21181         * config/rs6000/singlefp.h: Delete.
21182         * config/rs6000/sysv4.h: Remove Xilinx FP support.
21183         * config/rs6000/t-rs6000: Ditto.
21184         * config/rs6000/t-xilinx: Delete.
21185         * config/rs6000/titan.md: Adjust for fp_type removal.
21186         * config/rs6000/vsx.md: Remove Xilinx FP support.
21187         (VStype_simple): Delete.
21188         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
21189         * config/rs6000/xfpu.h: Delete.
21190         * config/rs6000/xfpu.md: Delete.
21191         * config/rs6000/xilinx.h: Delete.
21192         * config/rs6000/xilinx.opt: Delete.
21193         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
21194         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
21196 2018-05-04  Tom de Vries  <tom@codesourcery.com>
21198         PR libgomp/85639
21199         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
21200         if ignore == 0.
21202 2018-05-04  Richard Biener  <rguenther@suse.de>
21204         PR middle-end/85627
21205         * tree-complex.c (update_complex_assignment): We are always in SSA form.
21206         (expand_complex_div_wide): Likewise.
21207         (expand_complex_operations_1): Likewise.
21208         (expand_complex_libcall): Preserve EH info of the original stmt.
21209         (tree_lower_complex): Handle removed blocks.
21210         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
21211         on complex multiplication and division libcall builtins.
21213 2018-05-04  Richard Biener  <rguenther@suse.de>
21215         PR middle-end/85574
21216         * fold-const.c (negate_expr_p): Restrict negation of operand
21217         zero of a division to when we know that can happen without
21218         overflow.
21219         (fold_negate_expr_1): Likewise.
21221 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
21223         PR libstdc++/85466
21224         * real.h (real_nextafter): Declare.
21225         * real.c (real_nextafter): New function.
21226         * fold-const-call.c (fold_const_nextafter): New function.
21227         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
21228         CASE_CFN_NEXTTOWARD.
21229         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
21230         even when arg1_mode is different from arg0_mode.
21232 2018-05-03  Nathan Sidwell  <nathan@acm.org>
21234         * doc/extend.texi (Deprecated Features): Remove
21235         -ffriend-injection.
21236         (Backwards Compatibility): Likewise.
21237         * doc/invoke.texi (C++ Language Options): Likewise.
21238         (C++ Dialect Options): Likewise.
21240 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
21242         PR target/85530
21243         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
21244         _mm512_mask_mullox_epi64): New intrinsics.
21246 2018-05-03  Tom de Vries  <tom@codesourcery.com>
21248         PR testsuite/85106
21249         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21250         dump files): Add offload-tree.
21252 2018-05-03  Richard Biener  <rguenther@suse.de>
21254         PR tree-optimization/85615
21255         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
21256         to loops not nested in BBs loop father to avoid creating multi-entry
21257         loops.
21259 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21261         PR tree-optimization/70291
21262         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
21263         arguments.  Change return type to tree.  Emit libcall as a new
21264         statement rather than replacing existing one when inplace_p is true.
21265         (expand_complex_multiplication_components): New function.
21266         (expand_complex_multiplication): Expand floating-point complex
21267         multiplication using the above.
21268         (expand_complex_division): Rename inner_type parameter to type.
21269         Update expand_complex_libcall call-site.
21270         (expand_complex_operations_1): Update expand_complex_multiplication
21271         and expand_complex_division call-sites.
21273 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
21275         PR target/85582
21276         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
21277         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
21278         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
21279         the highest significant bit of the shift count mask is clear.  In
21280         check whether and[sq]i3 is needed verify that all significant bits
21281         of the shift count other than the highest are set.
21283 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21285         PR libgomp/82428
21286         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
21287         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
21288         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
21289         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
21290         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
21291         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
21292         __builtin_goacc_parlevel_size.
21294 2018-05-02  Richard Biener  <rguenther@suse.de>
21296         PR tree-optimization/85597
21297         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
21298         do not use split vect_get_vec_defs call but call vect_get_slp_defs
21299         directly.
21301 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21303         PR testsuite/85106
21304         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21305         dump files): Add ltrans-tree.
21307 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21309         PR testsuite/85106
21310         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21311         dump files): Add wpa-ipa.
21313 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
21315         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
21316         powerpc*-*-linux*paired* target.
21317         * config/rs6000/750cl.h: Delete.
21318         * config/rs6000/paired.h: Delete.
21319         * config/rs6000/paired.md: Delete.
21320         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
21321         float support.
21322         * config/rs6000/rs6000-builtin.def: Remove paired float support.
21323         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
21324         comment.  Remove paired float support.
21325         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
21326         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
21327         VECTOR_PAIRED.
21328         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
21329         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
21330         declarations.
21331         * config/rs6000/rs6000.c: Remove paired float support.
21332         (paired_expand_vector_init, paired_expand_vector_move,
21333         paired_emit_vector_compare, paired_emit_vector_cond_expr,
21334         (paired_expand_lv_builtin, paired_expand_stv_builtin,
21335         paired_expand_builtin, paired_expand_predicate_builtin,
21336         paired_init_builtins): Delete.
21337         * config/rs6000/rs6000.h: Remove paired float support.
21338         * config/rs6000/rs6000.md: Remove paired float support.
21339         (move_from_CR_ov_bit): Delete.
21340         * config/rs6000/rs6000.opt (mpaired): Delete.
21341         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
21342         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
21344 2018-05-02  Richard Biener  <rguenther@suse.de>
21346         PR middle-end/85567
21347         * gimplify.c (gimplify_save_expr): When in SSA form allow
21348         SAVE_EXPRs to compute to SSA vars.
21350 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
21352         PR target/85582
21353         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
21354         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
21355         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
21356         clobber operands[2], instead use a new pseudo.  Formatting fixes.
21358 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
21360         PR tree-optimization/85586
21361         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
21362         exit early for statements in the same group if the accesses are
21363         not strided.
21365 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21367         PR lto/85451
21368         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
21369         error message.
21371 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
21373         PR tree-optimization/85143
21374         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
21376 2018-05-01  Tom de Vries  <tom@codesourcery.com>
21378         PR lto/85451
21379         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
21380         not found" error message.
21382 2018-05-01  Tom de Vries  <tom@codesourcery.com>
21384         PR other/83786
21385         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
21386         * vec.c (test_ordered_remove_if): New function.
21387         (vec_c_tests): Call test_ordered_remove_if.
21388         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
21389         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
21390         * tree-vect-patterns.c (vect_pattern_recog_1): Use
21391         VEC_ORDERED_REMOVE_IF.
21393 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
21395         PR tree-optimization/82665
21396         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
21397         pointer subtraction where arguments come from a memchr call.
21399 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
21401         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
21402         --push-state --as-needed and --pop-state instead of --as-needed and
21403         --no-as-needed if ld supports it.
21404         * configure: Regenerated.
21406         PR web/85578
21407         * doc/install.texi2html: Replace _002d with - and _002a with * in
21408         generated html files using sed.
21410 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21412         PR c++/85523
21413         * gcc-rich-location.c (blank_line_before_p): New function.
21414         (use_new_line): New function.
21415         (gcc_rich_location::add_fixit_insert_formatted): New function.
21416         * gcc-rich-location.h
21417         (gcc_rich_location::add_fixit_insert_formatted): New function.
21419 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21421         * selftest.c (assert_streq): Rename "expected" and "actual" to
21422         "val1" and "val2".  Extend NULL-handling to cover both inputs
21423         symmetrically, while still requiring both to be non-NULL for a pass.
21424         * selftest.h (assert_streq): Rename "expected" and "actual" to
21425         "val1" and "val2".
21426         (ASSERT_EQ): Likewise.
21427         (ASSERT_EQ_AT): Likewise.
21428         (ASSERT_KNOWN_EQ): Likewise.
21429         (ASSERT_KNOWN_EQ_AT): Likewise.
21430         (ASSERT_NE): Likewise.
21431         (ASSERT_MAYBE_NE): Likewise.
21432         (ASSERT_MAYBE_NE_AT): Likewise.
21433         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
21434         the assertion to pass.
21435         (ASSERT_STREQ_AT): Likewise.
21437 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
21439         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
21440         interaction with -pie.
21442 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21444         * selftest.h: Fix alphabetization of per-source-file selftest
21445         declarations.
21447 2018-04-30  Jason Merrill  <jason@redhat.com>
21449         PR c++/61982 - dead stores to destroyed objects.
21450         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
21451         of clobber.
21453 2018-04-30  Jason Merrill  <jason@redhat.com>
21455         * tree.c (build_clobber): New.
21456         * tree.h: Declare it.
21457         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
21459 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21461         * diagnostic-show-locus.c (layout::layout): Update for
21462         location_get_source_line returning a char_span.
21463         (struct char_span): Move to input.h.
21464         (struct correction): Update for fields in char_span becoming
21465         private.
21466         (struct source_line): Update for location_get_source_line
21467         returning a char_span.
21468         (layout::print_line): Likewise.
21469         * edit-context.c (edited_file::print_content): Likewise.
21470         (edited_file::print_diff_hunk): Likewise.
21471         (edited_file::print_run_of_changed_lines): Likewise.
21472         (edited_file::get_num_lines): Likewise.
21473         (edited_line::edited_line): Likewise.
21474         * final.c (asm_show_source): Likewise.
21475         * input.c (location_get_source_line): Convert return type
21476         from const char * to char_span, losing the final "line_len"
21477         param.
21478         (dump_location_info): Update for the above.
21479         (get_substring_ranges_for_loc): Likewise.  Use a char_span
21480         when handling the literal within the line.
21481         (test_reading_source_line): Update for location_get_source_line
21482         returning a char_span.
21483         * input.h (class char_span): Move here from
21484         diagnostic-show-locus.c, converting from a struct to a class.
21485         Make data members private.
21486         (char_span::operator bool): New.
21487         (char_span::length): New.
21488         (char_span::get_buffer): New.
21489         (char_span::operator[]): New.
21490         (char_span::subspan): Make const.
21491         (char_span::xstrdup): New.
21492         (location_get_source_line): Convert return type from const char *
21493         to char_span, losing the final "line_size" param.
21495 2018-04-30  Jan Hubicka  <jh@suse.cz>
21497         * lto-wrapper.c (ltrans_priorities): New static var.
21498         (cmp_priority): New.
21499         (run_gcc): Read priorities and if doing parallel build order
21500         the Makefile by them.
21502 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21504         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
21506 2018-04-30  Richard Biener  <rguenther@suse.de>
21508         * tree-cfg.c (verify_address): Remove base argument, add
21509         flag whether to check TREE_ADDRESSABLE and do that.
21510         (verify_expr): Remove.
21511         (verify_types_in_gimple_reference): Add pieces from verify_expr.
21512         (verify_gimple_assign_single): Likewise.
21513         (verify_gimple_switch): Likewise.
21514         (verify_expr_location_1): Dereference tp once.  Add (disabled)
21515         piece from verify_expr.
21516         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
21518 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
21520         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
21522 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
21524         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
21525         (small_data_pattern): Likewise.
21526         (arc_rewrite_small_data): Likewise.
21527         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
21528         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
21529         (get_symbol_alignment): New function.
21530         (legitimate_small_data_address_p): Likewise.
21531         (legitimate_scaled_address): Update, call
21532         legitimate_small_data_address_p.
21533         (output_sdata): New static variable.
21534         (arc_print_operand): Update how we handle small data operands.
21535         (arc_print_operand_address): Likewise.
21536         (arc_legitimate_address_p): Update, use
21537         legitimate_small_data_address_p.
21538         (arc_rewrite_small_data_p): Remove.
21539         (arc_rewrite_small_data_1): Likewise.
21540         (arc_rewrite_small_data): Likewise.
21541         (small_data_pattern): Likewise.
21542         (compact_sda_memory_operand): Update to use
21543         legitimate_small_data_address_p and get_symbol_alignment.
21544         (prepare_move_operands): Don't rewite sdata pattern.
21545         (prepare_extend_operands): Remove.
21546         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
21547         pattern.
21548         (zero_extendqisi2): Likewise.
21549         (zero_extendhisi2): Likewise.
21550         (extendqihi2): Likewise.
21551         (extendqisi2): Likewise.
21552         (extendhisi2): Likewise.
21553         (addsi3): Likewise.
21554         (subsi3): Likewise.
21555         (andsi3): Likewise.
21556         * config/arc/constraints.md (Usd): Change it to memory constraint.
21558 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
21560         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
21561         as source of std instructions.
21562         * config/arc/arc.md (movsi_insn): Update pattern predicate to
21563         allow 6-bit constants as source for store instructions.
21564         (movdi_insn): Update instruction pattern to allow 6-bit constants
21565         as source for store instructions.
21567 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
21569         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
21571 2018-04-30  Nathan Sidwell  <nathan@acm.org>
21572             Sandra Loosemore <sandra@codesourcery.com>
21574         * dumpfile.c (dump_open): Allow '-' for stdout.
21575         * doc/invoke.texi (Developer Options): Document dump filename
21576         determination early.  Document stdin/stdout selection.
21578 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
21580         Microblaze Target: PIC data text relative
21582         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
21583         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
21584         Add declaration.
21585         * config/microblaze/microblaze.h (microblaze_constant_address_p):
21586         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
21587         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
21588         New addressing mode for data-text relative position indepenedent code.
21589         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
21590         'ADDRESS_SYMBOLIC_TXT_REL'.
21591         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
21592         (microblaze_legitimate_pic_operand): Exclude function calls from
21593         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
21594         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
21595         addresses cases.
21596         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
21597         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
21598         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
21599         for 'address + offset'.
21600         (microblaze_expand_prologue): Add new function prologue call for
21601         'r20' assignation.
21602         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
21603         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
21604         table in case of TARGET_PIC_DATA_TEXT_REL.
21605         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
21606         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
21607         Add new macros 'UNSPEC_TEXT',
21608         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
21609         + exclude function calls from 'UNSPEC_PLT' in case of data text
21610         relative mode.
21611         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21612         new target hook for generating address diff vector tables in case of
21613         flag_pic.
21614         * doc/tm.texi : Regenerate.
21615         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
21616         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
21617         of addr diff vector generation.
21618         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21619         target hook definition.
21620         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
21621         Add default function for generate_pic_addr_diff_vec -> flag_pic.
21622         * doc/invoke.texi (Add new pic option): Add new microblaze pic
21623         option for data text relative.
21625 2018-04-30  Richard Biener  <rguenther@suse.de>
21627         * tree-chrec.h (evolution_function_is_constant_p): Remove
21628         redundant check.
21629         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
21631 2018-04-30  Richard Biener  <rguenther@suse.de>
21633         PR bootstrap/85571
21634         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
21636 2018-04-30  Richard Biener  <rguenther@suse.de>
21638         PR tree-optimization/28364
21639         PR tree-optimization/85275
21640         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
21641         copying first exit test.
21643 2018-04-28  Mark Wielaard  <mark@klomp.org>
21645         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
21646         dwarf_version >= 5.
21647         (dwarf_AT): Handle DW_AT_addr_base.
21648         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
21650 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
21652         PR target/84431
21653         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
21654         (*ashl<dwi>3_doubleword_mask_1): Ditto.
21655         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
21656         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
21658 2018-04-28  Richard Biener  <rguenther@suse.de>
21660         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
21661         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
21662         to reflect use.  Only add interesting stmts.
21664 2018-04-27  Martin Jambor  <mjambor@suse.cz>
21666         PR ipa/85549
21667         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
21668         the jump function allows for passing through aggregate values.
21670 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
21672         * input.h (in_system_header_at): Convert from macro to inline
21673         function.
21674         (from_macro_expansion_at): Likewise.
21675         (from_macro_definition_at): Likewise.
21677 2018-04-27  Jeff Law  <law@redhat.com>
21679         * config.gcc: Mark tile* targets as deprecated/obsolete.
21681 2018-04-27  Richard Biener  <rguenther@suse.de>
21683         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21684         fix for ILP32.
21686 2018-04-27  Richard Biener  <rguenther@suse.de>
21688         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21690 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
21692         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21693         with Yd constraint. Set "preferred_for_speed" attribute from
21694         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21695         with Yd constraint.
21696         (*movdi_internal): Ditto.
21697         (movti_interunit splitters): Remove
21698         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21699         (movdi_interunit splitters): Ditto.
21700         * config/i386/constraints.md (Ye): Remove.
21701         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21703 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21705         PR target/85512
21706         * config/aarch64/constraints.md (Usg): Limit to 31.
21707         (Usj): Limit to 63.
21709 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
21711         PR tree-optimization/85529
21712         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21713         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21714         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21715         zero extension or masking of the MSB bit.
21716         (optimize_range_tests): Add FIRST_BB argument, pass it through
21717         to optimize_range_tests_var_bound.
21718         (maybe_optimize_range_tests, reassociate_bb): Adjust
21719         optimize_range_tests callers.
21721 2018-04-26  Richard Biener  <rguenther@suse.de>
21722             Jakub Jelinek  <jakub@redhat.com>
21724         * cgraph.h (symbol_table): Just declare debug method here.
21725         * symtab.c (symbol_table::debug): Define.
21727 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
21729         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21731 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
21733         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21734         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21735         (*movdi_internal): Substitute Yi and Yj constraint with x
21736         and Ym and Yn constraint with y constraint.  Update "isa"
21737         attribute and set "preferred_for_speed" attribute from
21738         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21739         (*movsi_internal): Ditto.
21740         (*movdf_internal): Ditto.
21741         (*movsf_internal): Ditto.
21742         (*zero_extendsidi2): Ditto.
21743         * config/i386/sse.md (vec_set<mode>_0): Ditto.
21744         (sse2_loadld): Ditto.
21745         (*vec_extract<ssevecmodelower>_0): Ditto.
21746         (*vec_extractv4si_0_zext_sse4): Ditto.
21747         (vec_concatv2di): Ditto.
21748         (*vec_dup<mode>): Ditto.
21749         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21750         * config/i386/constraints.md (Yi): Remove.
21751         (Yj): Remove.
21752         (Ym): Remove.
21753         (Yn): Remove.
21755 2018-04-26  Nathan Sidwell  <nathan@acm.org>
21757         * dumpfile.c (dump_open): New.
21758         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21759         (dump_finish): Detect stdio/stderr by value not name.
21761 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
21763         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21765 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21767         PR target/84952
21768         * config/nvptx/nvptx.c (verify_neutering_jumps)
21769         (verify_neutering_labels): New function
21770         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21772 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21774         PR target/84025
21775         * config/nvptx/nvptx.c (needs_neutering_p): New function.
21776         (nvptx_single): Use needs_neutering_p to skip over insns that do not
21777         need neutering.
21779 2018-04-26  Richard Biener <rguenther@suse.de>
21780             Tom de Vries  <tom@codesourcery.com>
21782         PR lto/85422
21783         * lto-streamer-out.c (output_function): Fixup loops if required to match
21784         discovery done in the reader.
21786 2018-04-26  Richard Biener  <rguenther@suse.de>
21788         PR tree-optimization/85116
21789         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21790         have a loop exit from the single latch predecessor.  Remove
21791         case of header with just condition.
21792         (ch_base::copy_headers): Exclude infinite loops from any
21793         processing.
21794         (pass_ch::execute): Record exits.
21796 2018-04-26  Richard Biener  <rguenther@suse.de>
21798         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21799         prologue cost vector and pass it to vect_get_load_cost.
21800         (vect_get_peeling_costs_all_drs): Likewise.
21801         (vect_peeling_hash_get_lowest_cost): Likewise.
21802         (vect_enhance_data_refs_alignment): Likewise.
21804 2018-04-26  Richard Biener  <rguenther@suse.de>
21806         PR middle-end/85450
21807         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21808         checking of integer<->pointer conversions.
21809         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21810         sign-/zero-extending pointer types.
21811         (expand_omp_for_static_chunk): Likewise.
21813 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21814             Jean Lee  <xiaoyur347@gmail.com>
21816         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21817         (TARGET_ASAN_SHADOW_OFFSET): Define.
21818         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21819         true for -fsanitize=address.
21821 2018-04-25  Mark Wielaard  <mark@klomp.org>
21823         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21824         shorter ones.
21826 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21828         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21829         than "alu", remove explicit "memory" and "imm_disp" attributes.
21830         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21832         PR middle-end/85414
21833         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21834         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21835         gen_lowpart_no_emit.
21837 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21839         PR target/85473
21840         * config/i386/i386.c (ix86_expand_builtin): Change memory
21841         operand to XI, extend p0 to Pmode.
21842         * config/i386/i386.md: Change unspec volatile and operand
21843         1 mode to XI, change operand 0 mode to P.
21845 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21847         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21848         GET_MODE_MASK before any checking.
21849         (nds32_can_use_bset_p): Likewise.
21850         (nds32_can_use_btgl_p): Likewise.
21852 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21854         * config/nds32/nds32-doubleword.md: New define_split pattern for
21855         illegal register number.
21857 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21859         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21861 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21863         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21865 2018-04-25  Richard Biener  <rguenther@suse.de>
21867         * lto-streamer.h (LTO_major_version): Bump to 8.
21869 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21871         * BASE-VER: Set to 9.0.0.
21873 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21875         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21876         in __abskf2 and __powikf2.
21878 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21880         PR target/85512
21881         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21882         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21883         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21884         Use the above on operand 2.  Reindent.
21885         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21887 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21889         PR target/85485
21890         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21891         handle OPT_mcet.
21892         * config/i386/i386.opt (mcet): Removed.
21893         * doc/install.texi: Remove -mcet documentation.
21894         * doc/invoke.texi: Likewise.
21896 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21898         PR target/85485
21899         * doc/install.texi: Remove -mcet from bootstrap-cet.
21901 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21903         PR target/85511
21904         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21905         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21906         if TARGET_64BIT.
21908         PR target/85503
21909         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21910         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21911         containing a CONST_VECTOR.
21913 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21915         * doc/install.texi: Update newlib dependency for nvptx.
21917 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21919         PR target/85508
21920         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21921         instead of INTVAL when shifting x left.
21923 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21925         PR tree-optimization/85478
21926         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21927         vect_grouped_store_supported for single element vectors.
21929 2018-04-24  Richard Biener  <rguenther@suse.de>
21931         PR target/85491
21932         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21933         load cost increase to the case of non-constant step.
21935 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21937         PR target/84828
21938         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21939         destination if any_malformed_asm.
21941 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21943         PR middle-end/85496
21944         * expr.c (store_field): In the bitfield case, if the value comes from
21945         a function call and is returned in registers by means of a PARALLEL,
21946         do not change the mode of the temporary unless BLKmode and VOIDmode.
21948 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21950         PR rtl-optimization/85423
21951         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21952         dependencies to debug insns when the previous insn is non-debug.
21954 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21956         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21957         enums into a single definition.
21958         (fls): Fix predicates and printing.
21959         (seti): Likewise.
21961 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21963         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21964         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21965         and short u6 immediate.
21966         (check_if_valid_sleep_operand): Remove.
21967         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21969 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21971         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21972         flag_always_save_lp condition.
21973         * config/nds32/nds32.opt (malways-save-lp): New option.
21975 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21977         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21978         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21979         * config/nds32/nds32.h
21980         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21981         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21983 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21985         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21986         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21988 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21989             Chung-Ju Wu  <jasonwucj@gmail.com>
21991         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21992         Declare.
21993         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21994         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21996 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21998         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
22000 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22002         * config/nds32/nds32-protos.h (nds32_data_alignment,
22003         nds32_local_alignment): Declare.
22004         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
22005         nds32_local_alignment): New functions.
22006         (TARGET_CONSTANT_ALIGNMENT): Define.
22007         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
22009 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22011         * config/nds32/nds32.c
22012         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
22013         (TARGET_MODES_TIEABLE_P): Likewise.
22015 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22017         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
22018         level Ofast and Og.
22020 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
22021             Chung-Ju Wu  <jasonwucj@gmail.com>
22023         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22024         for unaligned access.
22025         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22026         unaligned access.
22027         * config/nds32/nds32-intrinsic.md: Likewise.
22028         * config/nds32/nds32_intrinsic.h: Likewise.
22029         * config/nds32/nds32.h (nds32_builtins): Likewise.
22030         * config/nds32/nds32.opt (munaligned-access): New option.
22031         * config/nds32/nds32.c (nds32_asm_file_start): Display
22032         flag_unaligned_access status.
22034 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
22036         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
22037         -mno-relax is present.
22038         * config/riscv/linux.h (LINK_SPEC): Ditto.
22040 2018-04-20  Martin Sebor  <msebor@redhat.com>
22042         PR c/85365
22043         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
22044         for null pointers.
22045         (gimple_fold_builtin_stxcpy_chk): Same.
22046         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
22048 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
22050         PR target/85456
22051         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
22052         __powikf2 when long double is IEEE 128-bit.
22054 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
22056         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
22057         step to make sure stack always aligned.
22059 2018-04-20  Carl Love  <cel@us.ibm.com>
22061         PR target/83402
22062         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
22063         size check for arg0.
22065 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
22066             Tom de Vries  <tom@codesourcery.com>
22068         PR target/85445
22069         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
22070         Emit insns for calls too.
22071         (nvptx_find_par): Always look for worker-level predecessor insn.
22072         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
22073         calls.
22074         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
22075         (nvptx_process_pars): Propagate frames for calls.
22077 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
22079         PR target/85469
22080         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
22081         Removed.
22082         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
22083         (ix86_handle_option): Don't handle OPT_mibt.
22084         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
22085         __SHSTK__.
22086         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
22087         has_ibt and ibt.
22088         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
22089         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
22090         (ix86_target_macros): Define __CET__ with flag_cf_protection
22091         for -fcf-protection.
22092         * config/i386/i386.c (isa2_opts): Remove -mibt.
22093         * config/i386/i386.h (TARGET_IBT): Removed.
22094         (TARGET_IBT_P): Likewise.
22095         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
22096         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
22097         * config/i386/i386.opt (mcet): Update help message.
22098         (mshstk): Likewise.
22099         (mibt): Removed.
22100         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
22101         -mcet as an alias for -mshstk.
22103 2018-04-20  Richard Biener <rguenther@suse.de>
22105         PR middle-end/85475
22106         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
22107         complexity by forcing a single use of the multiply operand.
22109 2018-04-20  Martin Jambor  <mjambor@suse.cz>
22111         ipa/85449
22112         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
22113         recursion dependency to only apply to non-clones.
22115 2018-04-20  Martin Jambor  <mjambor@suse.cz>
22117         ipa/85447
22118         * ipa-cp.c (create_specialized_node): Check that clones of
22119         self-recursive edges exist during IPA-CP.
22121 2018-04-19  Toon Moene  <toon@moene.org>
22123         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
22124         by -O3.
22126 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
22128         PR tree-optimization/85467
22129         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
22130         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
22131         VECTOR_CST element to type.
22133 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22135         PR target/85397
22136         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
22137         * config/i386/i386.md (builtin_setjmp_setup): Removed.
22138         (builtin_longjmp): Likewise.
22139         (save_stack_nonlocal): New pattern.
22140         (restore_stack_nonlocal): Likewise.
22142 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22144         PR target/85404
22145         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22146         Replace ASM_OUTPUT_LABEL with fprintf.
22148 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22150         PR target/85417
22151         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22152         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
22153         * config/i386/i386-c.c (ix86_target_macros_internal): Also
22154         define __IBT__ and __SHSTK__ for -fcf-protection.
22155         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
22156         TARGET_IBT.
22157         (ix86_trampoline_init): Likewise.
22158         (x86_output_mi_thunk): Likewise.
22159         (ix86_notrack_prefixed_insn_p): Likewise.
22160         (ix86_option_override_internal): Don't disallow -fcf-protection.
22161         * config/i386/i386.md (rdssp<mode>): Also enable for
22162         -fcf-protection.
22163         (incssp<mode>): Likewise.
22164         (nop_endbr): Likewise.
22165         * config/i386/i386.opt (mcet): Change help message to built-in
22166         functions only.
22167         (mibt): Likewise.
22168         (mshstk): Likewise.
22169         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
22170         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
22171         enable CET built-in functions.
22173 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
22175         * common/config/i386/i386-common.c
22176         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
22177         OPTION_MASK_ISA_MOVDIRI_UNSET,
22178         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
22179         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
22180         * config.gcc (movdirintrin.h): New header.
22181         * config/i386/cpuid.h (bit_MOVDIRI,
22182         bit_MOVDIR64B): New bits.
22183         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
22184         and -mmvodir64b.
22185         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
22186         (VOID, PVOID, PCVOID)): New function types.
22187         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
22188         __builtin_ia32_directstoreu_u64,
22189         __builtin_ia32_movdir64b): New builtins.
22190         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
22191         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
22192         and -mmovdiri.
22193         (ix86_valid_target_attribute_inner_p): Ditto.
22194         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
22195         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
22196         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
22197         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
22198         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
22199         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
22200         (movdiri<mode>, movdir64b_<mode>): New.
22201         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
22202         * config/i386/immintrin.h: Include movdirintrin.h.
22203         * config/i386/movdirintrin.h: New file.
22204         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
22206 2018-04-19  Richard Biener  <rguenther@suse.de>
22208         PR middle-end/85455
22209         * cfg.c (clear_bb_flags): When loop state says we have
22210         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
22212 2018-04-19  Richard Biener  <rguenther@suse.de>
22214         PR tree-optimization/84737
22215         * tree-vect-data-refs.c (vect_copy_ref_info): New function
22216         copying restrict info.
22217         (vect_setup_realignment): Use it.
22218         * tree-vectorizer.h (vect_copy_ref_info): Declare.
22219         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
22220         the first DR to all generated stores.
22221         (vectorizable_load): Likewise for loads.
22223 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
22225         PR tree-optimization/85446
22226         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
22227         the integral and pointer types to have the same precision.
22229         * doc/install.texi: Document --disable-cet being the default and
22230         --enable-cet=auto.
22232 2018-04-18  Martin Liska  <mliska@suse.cz>
22234         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
22235         style.
22237 2018-04-18  Martin Liska  <mliska@suse.cz>
22239         Revert
22240         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22242         PR ipa/83983
22243         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22244         arguments if they are comparable.
22246 2018-04-18  Martin Liska  <mliska@suse.cz>
22248         Revert
22249         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
22251         PR lto/84805
22252         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
22253         incomplete types.
22255 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
22257         PR target/85388
22258         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
22259         ENDBR after calling __morestack.
22261 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
22263         PR jit/85384
22264         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
22265         by using gcc_base_ver to generate a gcc_driver_version, and use
22266         it when generating GCC_DRIVER_NAME.
22267         * configure: Regenerate.
22269 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
22271         PR target/81084
22272         * config.gcc: Obsolete powerpc*-*-*spe*.
22274 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22276         PR debug/84637
22277         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
22278         (stabstr_D): Change type of unum from unsigned int to
22279         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
22280         type.
22282 2018-04-17  Jim Wilson  <jimw@sifive.com>
22284         PR 84856
22285         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
22286         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
22287         Set arg_pointer_offset after using pretend_args_size.
22289 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22291         PR rtl-optimization/85431
22292         * dse.c (record_store): Ignore zero width stores.
22294         PR sanitizer/85230
22295         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
22296         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
22297         __builtin_stack_restore rather than after it.
22298         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
22299         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
22300         argument instead of virtual_dynamic_stack_rtx.
22302 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22304         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
22305         New prototype.
22306         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22307         Add note to error message to explain internal mapping of overloaded
22308         built-in function name to non-overloaded built-in function name.
22309         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
22310         function.
22312 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
22314         PR target/85424
22315         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
22316         where the inputs overlap with the output.
22318 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22320         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
22321         (=v, v) alternative and explicit "memory" attribute.
22322         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
22323         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22324         attributes.
22325         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
22326         "sselog1" type instead of "sselog".
22327         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
22328         "sselog".  Remove explicit "memory" attribute.
22329         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
22330         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22331         attributes.
22332         (vec_extract_hi_v32hi): Merge all alternatives into one, use
22333         "sselog1" type instead of "sselog".  Remove explicit "memory"
22334         attribute.
22335         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
22336         use "sselog1" type instead of "sselog".  Remove explicit "memory"
22337         attribute.
22338         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
22339         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22340         attributes.
22341         (vec_extract_hi_v64qi): Merge all alternatives into one, use
22342         "sselog1" type instead of "sselog".  Remove explicit "memory"
22343         attribute.
22344         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
22345         use "sselog1" type instead of "sselog".  Remove explicit "memory"
22346         attribute.
22348         PR target/85430
22349         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
22351         PR middle-end/85414
22352         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
22353         on a SUBREG.
22355 2018-04-17  Martin Jambor  <mjambor@suse.cz>
22357         PR ipa/85421
22358         * ipa-cp.c (create_specialized_node): Call
22359         expand_all_artificial_thunks if necessary.
22361 2018-04-17  Martin Liska  <mliska@suse.cz>
22363         PR lto/85405
22364         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
22365         in message, remote space in between '_G' and '('.
22367 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22369         PR target/85281
22370         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
22371         avx512f_vmcmp<mode>3<round_saeonly_name>,
22372         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
22373         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
22374         avx512f_rndscale<mode><round_saeonly_name>,
22375         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
22376         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
22377         Use %<iptr>2 instead of %2 for -masm=intel.
22378         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
22379         avx512f_vcvttss2usi<round_saeonly_name>,
22380         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
22381         -masm=intel.
22382         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
22383         avx512f_vcvttsd2usi<round_saeonly_name>,
22384         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
22385         Use %q1 instead of %1 for -masm=intel.
22386         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
22387         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
22388         of %3 for -masm=intel.
22389         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
22390         -masm=intel.
22391         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
22392         -masm=intel.
22393         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
22394         -masm=intel.
22395         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
22396         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
22397         %g1.
22398         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
22399         -masm=intel.
22400         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
22401         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
22402         %g1 and one with %0 and %1.
22403         (avx512er_vmrcp28<mode><round_saeonly_name>,
22404         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
22405         %1 for -masm=intel.
22406         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
22407         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
22408         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
22409         of %0 and %{%4%} for -masm=intel.
22410         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
22411         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
22412         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
22413         order of %0 and %{%5%}%{z%} for -masm=intel.
22415 2018-04-17  Jan Hubicka  <jh@suse.cz>
22417         PR lto/85405
22418         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
22420 2018-04-17  Martin Liska  <mliska@suse.cz>
22422         PR ipa/85329
22423         * multiple_target.c (create_dispatcher_calls): Set apostrophes
22424         for target_clone error message.  Make default implementation
22425         clone to be a local declaration.
22426         (separate_attrs): Add new argument and check for an empty
22427         string.
22428         (expand_target_clones): Handle it.
22429         (ipa_target_clone): Make redirection just for target_clones
22430         functions.
22432 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
22433             Tom de Vries  <tom@codesourcery.com>
22435         PR middle-end/84955
22436         * omp-expand.c (expand_oacc_for): Add dummy false branch for
22437         tiled basic blocks without omp continue statements.
22439 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
22441         PR target/83660
22442         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
22443         vec_extract expression as having side effects to make sure it gets
22444         a cleanup point.
22446 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
22448         PR target/85403
22449         * config/i386/i386.c (get_builtin_code_for_version): Check
22450         error_mark_node.
22452 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
22454         PR target/84331
22455         * config.gcc: Support "skylake".
22456         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22457         PROCESSOR_SKYLAKE.
22458         * config/i386/i386.c (m_SKYLAKE): Define.
22459         (processor_target_table): Add "skylake".
22460         (ix86_option_override_internal): Add "skylake".
22461         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
22462         PROCESSOR_CANNONLAKE.
22463         (get_builtin_code_for_version): Fix priority for
22464         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
22465         PROCESSOR_SKYLAKE-AVX512.
22466         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
22467         (processor_type): Add PROCESSOR_SKYLAKE.
22469 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
22470             Jason Merrill  <jason@redhat.com>
22472         PR c++/85112
22473         * convert.c (convert_to_integer_1): Use direct recursion for
22474         enumeral types and types with a precision less than the number
22475         of bits in their mode.
22477 2018-04-16  Julia Koval  <julia.koval@intel.com>
22479         PR target/84413
22480         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
22481         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
22483 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
22485         PR target/85293
22486         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
22487         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
22488         and -mno-direct-move.
22490 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
22492         PR target/83402
22493         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
22494         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
22495         Ensure negative shifts result in {0}.
22497 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
22499         PR rtl-optimization/79916
22500         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
22501         regs (if any) to define how to gnerate SD moves when LRA is in
22502         progress.
22504 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
22506         PR rtl-optimization/85393
22507         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
22508         * except.c (expand_dw2_landing_pad_for_region): Make static.
22509         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
22510         a label and unconditional jump to old_bb, rather than
22511         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
22512         basic block.
22514         PR rtl-optimization/85376
22515         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
22516         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
22517         instead of a specific value.
22519 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
22520             Bin Cheng  <bin.cheng@arm.com>
22522         PR tree-optimization/82965
22523         PR tree-optimization/83991
22524         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
22525         by_profile_only parameter.
22526         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
22527         information if the loop was predicted to iterate too many times.
22528         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
22530 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
22532         PR lto/71991
22533         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
22534         always inline.
22536 2018-04-13  Martin Liska  <mliska@suse.cz>
22537             Jakub Jelinek  <jakub@redhat.com>
22539         PR middle-end/81657
22540         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
22541         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
22542         * builtins.c (expand_builtin_memory_copy_args): Use
22543         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
22544         handle dest_addr == pc_rtx.
22546 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
22548         PR target/85291
22549         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
22550         asked to not generate direct moves.
22551         (fix_trunc<mode>si2_stfiwx): Similar.
22552         (fix_trunc<mode>si2_internal): Similar.
22554 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
22556         PR debug/83157
22557         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
22558         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
22559         lookup if dest in some wider mode is known to be const0_rtx and
22560         if so, record permanent equivalence for it to be ZERO_EXTEND of
22561         the narrower mode destination.
22563 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
22565         * lto-streamer-out.c (output_function): Revert 259346.
22566         * omp-expand.c (expand_oacc_for): Likewise.
22568 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
22570         PR rtl-optimization/85354
22571         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
22572         * sel-sched.c (sel_global_init): ... here.
22574 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
22576         PR target/85238
22577         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
22578         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
22579         mode for PE-COFF targets.
22580         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
22581         (i386_pe_asm_lto_end): Likewise.
22582         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
22583         (TARGET_ASM_LTO_END): Likewise.
22584         * config/i386/winnt.c (saved_debug_info_level): New static variable.
22585         (i386_pe_asm_lto_start): New function.
22586         (i386_pe_asm_lto_end): Likewise.
22588 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
22589             Richard Biener  <rguenther@suse.de>
22591         PR middle-end/84955
22592         * lto-streamer-out.c (output_function): Fix CFG loop state before
22593         streaming out.
22594         * omp-expand.c (expand_oacc_for): Handle calls to internal
22595         functions like regular functions.
22597 2018-04-12  Richard Biener  <rguenther@suse.de>
22599         PR lto/85371
22600         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
22601         for the early LTO debug to properly generate references to it
22602         during DIE emission.  Do not re-use that for the skeleton for
22603         split-dwarf.
22604         (dwarf2out_early_finish): Likewise.
22606 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
22608         PR target/85328
22609         * config/i386/sse.md
22610         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
22611         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
22612         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
22613         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
22614         and output is a reg, avoid creating invalid lowpart subreg, but
22615         instead split into a 512-bit move.  Don't split if not AVX512VL,
22616         input is xmm16+ reg and output is a mem.
22617         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
22618         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
22619         xmm16+ reg and output is a mem.
22621 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22623         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
22624         also for flag_dwarf2_cfi_asm.
22626 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
22628         PR rtl-optimization/85342
22629         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
22630         a bool scalar var inside of the loop instead.  Don't try to update
22631         recog_data.operand after failed apply_change_group.
22633 2018-04-12  Tom de Vries  <tom@codesourcery.com>
22635         PR target/85296
22636         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
22637         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
22638         array with flexible array member as array without given dimension.
22639         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
22640         argument for undefined param to true.
22642 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
22644         PR target/85321
22645         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22646         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
22647         from PowerPC section.
22648         * config/rs6000/sysv4.opt (mcall-): Improve help text.
22649         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
22650         help text that is too long.
22651         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
22652         help text that is too long.
22653         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
22654         help text that is too long.
22656 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
22658         * config/alpha/alpha.md (stack_probe_internal): Rename
22659         from "probe_stack".  Update all callers.
22661 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22663         PR rtl-optimization/84566
22664         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
22665         sched_macro_fuse_insns.
22667 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22669         PR target/84301
22670         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
22671         (compute_block_dependences): ... from here.
22673 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22675         PR tree-optimization/85331
22676         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
22677         from int to HOST_WIDE_INT.
22679 2018-04-11  Martin Jambor  <mjambor@suse.cz>
22681         PR ipa/84149
22682         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
22683         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22684         not the same as the source val.
22685         (cgraph_edge_brings_value_p): New parameter.
22686         (gather_edges_for_value): Pass destination value to
22687         cgraph_edge_brings_value_p.
22688         (perhaps_add_new_callers): Likewise.
22689         (get_info_about_necessary_edges): Likewise and exclude values brought
22690         only by self-recursive edges.
22691         (create_specialized_node): Redirect only clones of self-calling edges.
22692         (+self_recursive_pass_through_p): New function.
22693         (find_more_scalar_values_for_callers_subset): Use it.
22694         (find_aggregate_values_for_callers_subset): Likewise.
22695         (known_aggs_to_agg_replacement_list): Removed.
22696         (decide_whether_version_node): Re-calculate known constants for all
22697         remaining context clones.
22699 2018-04-11  Richard Biener  <rguenther@suse.de>
22701         PR lto/85339
22702         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22703         from early DWARF output.
22704         (dwarf2out_early_finish): Output line info unconditionally into
22705         early DWARF and add reference to it.
22707 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22709         PR target/85281
22710         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22711         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22712         other than V2DFmode using iptr mode attribute.
22713         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22715 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22717         PR rtl-optimization/84659
22718         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22720 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22722         PR debug/85302
22723         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22724         SIZEP is NULL.
22725         (output_loc_list): Pass address of a dummy size variable even in the
22726         locview handling loop.
22727         (index_location_lists): Add comment on why skip_loc_list_entry can't
22728         call size_of_locs.
22730 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22732         PR target/85261
22733         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22734         into register.
22736 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
22738         PR target/85321
22739         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22740         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22741         and -mstring-compare-inline-limit.
22743 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22745         PR target/85287
22746         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22747         for stack clash protection in a register whenever we need it to be in
22748         a register.
22750 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22752         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22753         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22755 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22757         PR target/85321
22758         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22759         the help text.
22760         (mlong-double-): Ditto.
22761         * config/rs6000/sysv4.opt (msdata=): Ditto.
22762         (mtls-size=): Ditto.
22764 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22766         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22767         erroneous entries for
22768         "vector int vec_ldl (int, long int *)", and
22769         "vector unsigned int vec_ldl (int, unsigned long int *)".
22770         Add comments and entries for
22771         "vector bool char vec_ldl (int, bool char *)",
22772         "vector bool short vec_ldl (int, bool short *)",
22773         "vector bool int vec_ldl (int, bool int *)",
22774         "vector bool long long vec_ldl (int, bool long long *)",
22775         "vector pixel vec_ldl (int, pixel *)",
22776         "vector long long vec_ldl (int, long long *)",
22777         "vector unsigned long long vec_ldl (int, unsigned long long *)".
22778         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22779         type tree bool_long_long_type_node and correct definition of
22780         bool_V2DI_type_node to make reference to this new type tree.
22781         (rs6000_mangle_type): Replace erroneous reference to
22782         bool_long_type_node with bool_long_long_type_node.
22783         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22784         comments to emphasize sign distinctions for char and int types and
22785         replace RS6000_BTI_bool_long constant with
22786         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
22787         use of RS6000_BTI_pixel.
22788         (bool_long_type_node): Remove this macro definition.
22789         (bool_long_long_type_node): New macro definition
22791 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22793         PR rtl-optimization/85300
22794         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22795         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22796         simplify_unary_operation fails.
22798 2018-04-10  Martin Liska  <mliska@suse.cz>
22800         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22801         cgraph_edge and ipa_ref.
22803 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22805         PR target/85177
22806         PR target/85255
22807         * config/i386/sse.md
22808         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22809         computation of the VEC_MERGE selector from mask.
22810         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22811         Fix decoding of the VEC_MERGE selector into mask.
22813 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22815         PR tree-optimization/85286
22816         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22818 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22820         * final.c (final_1): Set insn_last_address as well as
22821         insn_current_address.
22823 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22825         PR target/85173
22826         * explow.c (emit_stack_probe): Call validize_mem on memory location
22827         before passing it to gen_probe_stack.  Create address operand and
22828         legitimize it for the probe_stack_address case.
22830 2018-04-09  Jan Hubicka  <jh@suse.cz>
22832         PR lto/85078
22833         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22834         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22835         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22836         walk basetypes.
22837         (free_lang_data): Rebuild type inheritance graph.
22839 2018-04-09  Martin Sebor  <msebor@redhat.com>
22841         * invoke.texi (-finline-small-functions): Mention other optimization
22842         options.
22843         (-findirect-inlining, -fpartial-inlining): Same.
22844         (-finline-functions-called-once): Same.
22845         (-freorder-blocks-and-partition): Same.
22847 2018-04-09  Jan Hubicka  <jh@suse.cz>
22849         PR rtl/84058
22850         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22851         jumps; choose last target that matches the criteria (i.e.
22852         no partition changes for non-crossing jumps).
22853         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22854         support for redirecting crossing jumps to non-crossing.
22856 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22858         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22859         also for naked functions.
22861 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22863         * config/arc/arc.md (add_shift): New pattern.
22864         (add_shift2): Likewise.
22865         (sub_shift): Likewise.
22866         (sub_shift_cmp0_noout): Likewise.
22867         (compare_si_ashiftsi): Likewise.
22868         (xbfu_cmp0_noout): New combine pattern.
22869         (xbfu_cmp0"): Likewise.
22870         (movsi_set_cc_insn): Place the predicable variant first.
22871         (commutative_binary_cmp0_noout): Remove clobber.
22872         (commutative_binary_cmp0): New pattern.
22873         (noncommutative_binary_cmp0): Likewise.
22874         (noncommutative_binary_cmp0_noout): Likewise.
22875         (noncommutative_binary_comparison_result_used): Removed.
22876         (rsub_cmp0): New pattern.
22877         (rsub_cmp0_noout): Likewise.
22878         (extzvsi): Changed, keep only meaningful variants.
22879         (SQH, SEZ): New iterators.
22880         (SQH_postfix): New mode attribute.
22881         (SEZ_prefix): New code attribute.
22882         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22883         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22884         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22885         of numerical value.
22886         (noncommutative_operator): Check the availability of barrel
22887         shifter option.
22889 2018-04-09  Richard Biener  <rguenther@suse.de>
22891         PR tree-optimization/85284
22892         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22893         Only use the niter constraining form of simple_iv when the exit
22894         is always executed.
22896 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22898         PR target/84041
22899         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22900         (define_expand "*memory_barrier"): New define_expand.
22901         (define_insn "memory_barrier"): New insn.
22903 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22905         PR rtl-optimization/80463
22906         PR rtl-optimization/83972
22907         PR rtl-optimization/83480
22909         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22910         correct producer for the insn.
22911         (tidy_control_flow): Fixup seqnos in case of debug insns.
22913 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22915         PR rtl-optimization/83913
22917         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22918         different sched-times when merging exprs.
22920 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22922         PR rtl-optimization/83962
22924         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22925         tidy_fallthru_edge and tidy_control_flow.
22927 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22929         PR rtl-optimization/83530
22931         * sel-sched.c (force_next_insn): New global variable.
22932         (remove_insn_for_debug): When force_next_insn is true, also leave only
22933         next insn in the ready list.
22934         (sel_sched_region): When the region wasn't scheduled, make another pass
22935         over it with force_next_insn set to 1.
22937 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22939         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22940         into tm_file.
22941         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22942         for interrupt control.
22943         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22944         functions for interrupt control.
22945         * config/nds32/nds32-intrinsic.md: Likewise.
22946         * config/nds32/nds32_intrinsic.h: Likewise.
22947         * config/nds32/nds32.h (nds32_builtins): Likewise.
22949 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22951         * config/nds32/nds32.c (nds32_init_machine_status,
22952         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22953         strict_aligned_p field.
22954         (nds32_expand_to_rtl_hook): New function.
22955         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22956         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22958 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22959             Chung-Ju Wu  <jasonwucj@gmail.com>
22961         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22962         * config/nds32/nds32-n7.md: New file.
22963         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22964         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22965         pipeline.
22966         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22967         * config/nds32/nds32.md (pipeline_model): Add n7.
22968         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22969         * config/nds32/pipelines.md: Include n7 settings.
22971 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22972             Chung-Ju Wu  <jasonwucj@gmail.com>
22974         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22975         * config/nds32/nds32-e8.md: New file.
22976         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22977         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22978         pipeline.
22979         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22980         * config/nds32/nds32.md (pipeline_model): Add e8.
22981         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22982         * config/nds32/pipelines.md: Include e8 settings.
22984 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22985             Chung-Ju Wu  <jasonwucj@gmail.com>
22987         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22988         * config/nds32/nds32-n8.md: New file.
22989         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22990         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22991         pipeline.
22992         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22993         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22994         * config/nds32/nds32.md (pipeline_model): Add n8.
22995         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22996         * config/nds32/pipelines.md: Include n8 settings.
22998 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22999             Chung-Ju Wu  <jasonwucj@gmail.com>
23001         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
23002         * config/nds32/nds32-n9-2r1w.md: New file.
23003         * config/nds32/nds32-n9-3r2w.md: New file.
23004         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
23005         nds32_register_ports): New or modify for cpu n9.
23006         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
23007         pipeline.
23008         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
23009         * config/nds32/nds32-utils.c: New file.
23010         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
23011         TARGET_MUL_SLOW): Define.
23012         * config/nds32/nds32.md (pipeline_model): New attribute.
23013         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
23014         New options that support cpu n9.
23015         * config/nds32/pipelines.md: Include n9 settings.
23016         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
23018 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
23020         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
23021         information if necessary.
23022         (output_cond_branch_compare_zero): Likewise.
23023         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
23024         (nds32_target_alignment): Refine for alignment.
23025         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
23026         (FUNCTION_BOUNDARY): Modify.
23027         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
23028         align case.
23029         * config/nds32/nds32.opt (malways-align, malign-functions): New.
23031 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
23033         * config/nds32/constants.md (unspec_volatile_element): Add values for
23034         TLB operation and data prefetch.
23035         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
23036         functions for TLB operation and data prefetch.
23037         * config/nds32/nds32-intrinsic.md: Likewise.
23038         * config/nds32/nds32_intrinsic.h: Likewise.
23039         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
23040         (nds32_print_operand): Likewise.
23041         * config/nds32/nds32.h (nds32_builtins): Likewise.
23043 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
23044         Andrew Pinski <pinsika@gcc.gnu.org>
23046         PR middle-end/82976
23047         * match.pd: Use constant_boolean_node of correct type instead of
23048         boolean_true_node or boolean_false_node for simplifying
23049         pointer comparisons to zero.
23051 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
23053         PR tree-optimization/80021
23054         * tree.c (verify_type_variant): Make error call in verify_variant_match
23055         translatable and remove final full stop.
23057 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23059         * config/nds32/constants.md (unspec_volatile_element): Add
23060         UNSPEC_VOLATILE_EH_RETURN.
23061         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
23062         nds32_output_stack_pop): Support dwarf exception handling process.
23063         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
23064         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
23065         exception handling process.
23066         (nds32_compute_stack_frame): Likewise.
23067         (nds32_return_addr_rtx): Likewise.
23068         (nds32_initial_elimination_offset): Likewise.
23069         (nds32_expand_prologue): Likewise.
23070         (nds32_expand_epilogue): Likewise.
23071         (nds32_dynamic_chain_address): New function.
23072         * config/nds32/nds32.h (machine_function): Add fields for dwarf
23073         exception handling.
23074         (DYNAMIC_CHAIN_ADDRESS): Define.
23075         (EH_RETURN_DATA_REGNO): Define.
23076         (EH_RETURN_STACKADJ_RTX): Define.
23077         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
23078         patterns for dwarf exception handling.
23080 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23082         * config/nds32/nds32.h: Clean up obsolete macros.
23084 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23086         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23087         Add enum values for particular instructions.
23088         * config/nds32/nds32-intrinsic.c: Implementation of expanding
23089         particular intrinsic functions.
23090         * config/nds32/nds32-intrinsic.md: Likewise.
23091         * config/nds32/nds32_intrinsic.h: Likewise.
23092         * config/nds32/nds32.h (nds32_builtins): Likewise.
23093         * config/nds32/nds32.md (type): Add pbsad and pbsada.
23094         (btst, ave): New patterns for particular instructions.
23096 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23098         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23099         Add enum values for atomic load/store and memory sync.
23100         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
23101         and memory sync.
23102         * config/nds32/nds32-intrinsic.md: Likewise.
23103         * config/nds32/nds32_intrinsic.h: Likewise.
23104         * config/nds32/nds32.h (nds32_builtins): Likewise.
23106 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
23108         PR tree-optimization/85257
23109         * fold-const.c (native_encode_vector): If not all elts could fit
23110         and off is -1, return 0 rather than offset.
23111         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
23112         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
23113         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
23114         adjust buffer in native_interpret_expr call.
23116 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23118         * config/nds32/constants.md (unspec_volatile_element): Add cache
23119         control enum values.
23120         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
23121         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
23122         * config/nds32/nds32.c (nds32_cctl_names): New.
23123         (nds32_print_operand): Handle cache control register names.
23124         * config/nds32/nds32.h (nds32_builtins): New enum values.
23125         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
23126         macros.
23127         * config/nds32/nds32.md (type): Add mmu.
23128         * config/nds32/pipelines.md (simple_insn): Add mmu.
23130 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23132         * config/nds32/nds32.md (type): Remove call.
23133         * config/nds32/pipelines.md (simple_insn): Likewise.
23135 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23137         * config/nds32/constants.md (unspec_volatile_element): Add
23138         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
23139         UNSPEC_VOLATILE_FMFCFG.
23140         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
23141         description for fmfcfg and fmfcsr.
23142         (bdesc_1arg): Add fmtcsr.
23143         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
23144         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
23145         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
23146         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
23147         unspec_fmfcfg): New patterns.
23148         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
23149         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
23150         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
23151         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
23152         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
23153         __nds32__fmfcfg): Define.
23155 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23157         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
23158         intrinsic register names.
23159         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
23160         intrinsic register enum values and macros.
23162 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23164         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
23165         for load/store addressing form.
23166         (nds32_print_operand_address): Likewise.
23168 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
23170         PR target/85196
23171         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
23172         based on LABEL_REF.  Remove useless assertion.
23173         (pic_address_needs_scratch): Fix formatting.
23174         (sparc_legitimize_pic_address): Minor tweaks.
23175         (sparc_delegitimize_address): Adjust assertion accordingly.
23176         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
23177         into symbolic_operand.
23178         (movsi_high_pic_label_ref): Likewise.
23179         (movsi_lo_sum_pic_label_ref): Likewise.
23180         (movdi_pic_label_ref): Likewise.
23181         (movdi_high_pic_label_ref): Likewise.
23182         (movdi_lo_sum_pic_label_ref): Likewise.
23184 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
23186         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
23187         custom LIB_SPEC setup.
23189 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
23190             Kito Cheng  <kito.cheng@gmail.com>
23192         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
23193         * config/riscv/freebsd.h: New.
23195 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23197         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
23198         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
23199         file.
23201 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23202             Kito Cheng  <kito.cheng@gmail.com>
23204         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
23205         nds32_output_call, nds32_symbol_binds_local_p): New functions.
23206         * config/nds32/nds32-protos.h (nds32_output_call,
23207         nds32_output_return): Declare.
23208         * config/nds32/nds32.md: Refine all the call and return patterns.
23210 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
23212         PR debug/85252
23213         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
23214         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
23216         PR rtl-optimization/84872
23217         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
23218         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
23219         EDGE_CROSSING edge.
23221 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
23223         * expr.c (copy_blkmode_to_reg): Revert 254862.
23224         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
23226 2018-04-06  Richard Biener  <rguenther@suse.de>
23228         PR middle-end/85244
23229         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
23230         after seeing a component reference with an adjacent field.  Treat
23231         refs to arrays at struct end of external decls similar to
23232         refs to unconstrained commons.
23234 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
23236         PR sanitizer/85213
23237         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
23238         look through SAVE_EXPRs with non-side-effects argument.  Adjust
23239         recursive calls.
23240         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
23241         save_p here.
23243 2018-04-06  Richard Biener  <rguenther@suse.de>
23245         PR middle-end/85180
23246         * alias.c (find_base_term): New wrapper around find_base_term
23247         unwinding CSELIB_VAL_PTR changes.
23248         (find_base_term): Do not restore CSELIB_VAL_PTR during the
23249         recursion.
23251 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23253         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
23254         instructions.
23255         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
23256         constant definitions.
23257         ("nop"): lr 0,0 -> nopr r0
23258         ("nop_lr0", "nop_lr1"): New insn definitions.
23260 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23262         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
23263         NDS32_V3PUSH_AVAILABLE_P macro.
23265 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
23266             Chung-Ju Wu  <jasonwucj@gmail.com>
23268         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
23269         (nds32*-*-*): Add float and fpu_config into supported_defaults.
23270         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
23271         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
23272         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
23273         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
23274         * config/nds32/constraints.md: New constraints and checking for hard
23275         float configuration.
23276         * config/nds32/iterators.md: New mode iterator and attribute for hard
23277         float configuration.
23278         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
23279         patterns.
23280         * config/nds32/nds32-fpu.md: New file.
23281         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
23282         deal with hard float code generation.
23283         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
23284         ARCH_V3S.
23285         (abi_type, float_reg_number): New enum type.
23286         * config/nds32/nds32-predicates.c: New predicates for hard float.
23287         * config/nds32/nds32-protos.h: Declare functions for hard float.
23288         * config/nds32/nds32.c: Implementation for hard float configuration.
23289         * config/nds32/nds32.h: Definitions for hard float configuration.
23290         * config/nds32/nds32.md: Include hard float machine description and
23291         modify patterns for hard float configuration.
23292         * config/nds32/nds32.opt: New options for hard float configuration.
23293         * config/nds32/predicates.md: New predicates for hard float
23294         configuration.
23296 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
23298         * common/config/nds32/nds32-common.c
23299         (nds32_option_optimization_table): Enable -mreleax-hint by default.
23301 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
23303         PR middle-end/85195
23304         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
23305         CONSTRUCTOR_ELT (ctor, ...)->value.
23307 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
23309         PR target/85193
23310         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
23312 2018-04-05  Tom de Vries  <tom@codesourcery.com>
23314         PR target/85204
23315         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
23316         cond jump.
23318 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
23319             Kito Cheng  <kito.cheng@gmail.com>
23321         * config/nds32/constraints.md (U33): Fine-tune checking condition.
23322         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
23323         * config/nds32/nds32.h (nds32_16bit_address_type): Add
23324         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
23326 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
23327             Kito Cheng  <kito.cheng@gmail.com>
23329         * config/nds32/constraints.md (Ufe): New memory constraint.
23330         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
23331         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
23332         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
23333         operands.
23334         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
23335         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
23337 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23339         * config/nds32/nds32.md: Use optimize_size in the condition for
23340         alu-shift instructions.
23342 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23344         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
23346 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23348         * config/nds32/nds32.md (negsi2): Refine pattern.
23350 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
23351             Chung-Ju Wu  <jasonwucj@gmail.com>
23353         * config/nds32/iterators.md (shift_rotate): New code iterator.
23354         (shift): New code attribute.
23355         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
23356         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
23357         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
23358         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
23359         bit-wise operations.
23360         (andsi3, *andsi3): Ditto.
23361         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
23362         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
23363         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
23364         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
23365         nds32_ior_operand, nds32_xor_operand): New predicates.
23367 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23369         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
23370         (addsi3, subsi3): ... this.
23372 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23374         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
23376 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23378         * config/nds32/nds32.md: Adjust indention.
23380 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
23382         * config/nds32/nds32.md (feature): New attribute.
23384 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23386         * config/nds32/nds32.md (subtype): New attribute.
23388 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
23390         PR target/85203
23391         * config/arm/arm-builtins.c (arm_expand_builtin): Change
23392         expansion to perform a bitwise AND of the argument followed by a
23393         boolean negation of the result.
23395 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
23397         PR rtl-optimization/84878
23398         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
23399         the basic block.  Assert the use reference is not artificial and that
23400         it has an associated insn.
23402 2018-04-04  Michael Matz  <matz@suse.de>
23404         * builtins.c (compute_objsize): Pass correct operand
23405         to array_at_struct_end_p.
23407 2018-04-04  Richard Biener  <rguenther@suse.de>
23409         PR lto/85176
23410         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
23411         from contexts for DINFO_LEVEL_TERSE and below.
23413 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23415         * config/nds32/nds32-doubleword.md (move_<mode>): Require
23416         resiter_operand condition.
23417         * config/nds32/nds32.md (*move<mode>): Ditto.
23419 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23420             Monk Chiang  <sh.chiang04@gmail.com>
23422         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
23424 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23426         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
23428 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23429             Kito Cheng  <kito.cheng@gmail.com>
23431         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
23432         nds32_cond_code_str, output_cond_branch,
23433         output_cond_branch_compare_zero, nds32_expand_cbranch,
23434         nds32_expand_cstore, nds32_expand_movcc,
23435         nds32_output_cbranchsi4_equality_zero,
23436         nds32_output_cbranchsi4_equality_reg,
23437         nds32_output_cbranchsi4_equality_reg_or_const_int,
23438         nds32_output_cbranchsi4_greater_less_zero: New functions.
23439         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
23440         nds32_expand_cstore, nds32_expand_movcc,
23441         nds32_output_cbranchsi4_equality_zero,
23442         nds32_output_cbranchsi4_equality_reg,
23443         nds32_output_cbranchsi4_equality_reg_or_const_int,
23444         nds32_output_cbranchsi4_greater_less_zero): Declare.
23445         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
23446         nds32_rimm11s_operand): New predicates.
23447         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
23448         * config/nds32/nds32.md: Rewrite all the branch and conditional move
23449         patterns.
23451 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23453         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
23454         * config/nds32/nds32.md: Ditto.
23455         * config/nds32/pipelines.md: Ditto.
23457 2018-04-04  Richard Biener  <rguenther@suse.de>
23459         PR tree-optimization/85168
23460         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
23461         propagating abnormals.
23463 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23465         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
23467 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23468             Kito Cheng  <kito.cheng@gmail.com>
23470         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
23471         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
23472         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
23473         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
23474         * config/nds32/nds32.md (sibcall_internal): New.
23475         (sibcall_register): Remove.
23476         (sibcall_immediate): Remove.
23477         (sibcall_value_internal): New.
23478         (sibcall_value_register): Remove.
23479         (sibcall_value_immediate): Remove.
23480         * config/nds32/predicates.md (nds32_general_register_operand): New.
23481         (nds32_call_address_operand): New.
23483 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
23485         PR rtl-optimization/85167
23486         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
23487         bb_defs if *split_p, instead preinitialize it to NULL.
23489         PR tree-optimization/85156
23490         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
23491         evaluating the argument multiple times.
23493 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
23495         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
23496         than vector.
23497         (_mm_cvtpd_ps): Likewise.
23498         (_mm_cvttpd_epi32): Likewise.
23499         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
23500         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
23501         vector, pixel, and bool following altivec.h include.
23503 2018-04-03  Martin Sebor  <msebor@redhat.com>
23505         * doc/extend.texi (Common Function Attributes): Clarify.
23506         (const attribute): Likewise.
23507         (pure attribute): Likewise.
23509 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
23511         PR target/85169
23512         * config/i386/i386.c (ix86_expand_vector_set): Use
23513         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
23515 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
23517         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
23518         instructions when changing rounding bits to preserve precision bits
23519         in the x87 control word.
23521 2018-04-03  Martin Liska  <mliska@suse.cz>
23523         PR tree-optimization/82491
23524         * rtl.h (strip_offset_and_add): Replace += suboffset with
23525         poly_uint64 () + suboffset.
23527 2018-03-29  Martin Liska  <mliska@suse.cz>
23528             Martin Jambor  <mjambor@suse.cz>
23530         PR ipa/84947
23531         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
23532         param_type is not an integral or pointer type.
23534 2018-04-03  Richard Biener  <rguenther@suse.de>
23536         * sese.h (recompute_all_dominators): Remove.
23538 2018-04-02  Martin Sebor  <msebor@redhat.com>
23540         * doc/invoke.texi (-Wrestrict): Fix typos.
23542 2018-04-02  Jim Wilson  <jimw@sifive.com>
23544         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
23545         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
23546         (<optab>di3, <optab>si3_extend): Likewise.
23547         (<optab>si3_mask, <optab>si3_mask_1): New.
23548         (<optab>di3_mask, <optab>di3_mask_1): New.
23549         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
23550         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
23551         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
23553 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
23555         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
23556         example.
23558 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
23560         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
23561         (nds32_canonicalize_comparison): New function.
23563 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
23564             Kito Cheng  <kito.cheng@gmail.com>
23565             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
23567         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
23568         * config/nds32/constants.md (unspec_volatile_element): Add
23569         UNSPEC_VOLATILE_RELAX_GROUP.
23570         * config/nds32/nds32-relax-opt.c: New file.
23571         * config/nds32/nds32-predicates.c
23572         (nds32_symbol_load_store_p): New function.
23573         * config/nds32/nds32-protos.h
23574         (nds32_symbol_load_store_p): Declare function.
23575         (make_pass_nds32_relax_opt): Declare new rtl pass function.
23576         * config/nds32/nds32.c
23577         (nds32_register_pass): New function to register pass.
23578         (nds32_register_passes): New function to register passes.
23579         * config/nds32/nds32.md (relax_group): New pattern.
23580         * config/nds32/nds32.opt (mrelax-hint): New option.
23581         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
23583 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
23585         * config/nds32/t-nds32: Modify files dependency.
23587 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
23589         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
23590         (PROFILE_HOOK): Define its implementation.
23592 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
23594         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
23595         type and 32-bit size.
23597 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
23599         PR middle-end/85090
23600         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
23601         (V_128_256): New mode iterator.
23602         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
23603         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
23604         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
23605         of V.
23606         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
23607         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
23609 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
23611         PR target/83315
23612         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
23613         NaN inputs correctly.
23615 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
23617         PR target/80546
23618         * config/rs6000/vsx.md (??r): New mode attribute.
23619         (*vsx_mov<mode>_64bit): Use it.
23620         (*vsx_mov<mode>_32bit): Likewise.
23622 2018-03-30  Martin Sebor  <msebor@redhat.com>
23624         PR tree-optimization/84818
23625         * builtins.c (check_access): Use warning_n.
23627 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23629         PR target/83822
23630         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
23631         condition.
23632         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
23633         condition.
23635 2018-03-30  Julia Koval  <julia.koval@intel.com>
23637         PR target/84413
23638         * x86-tune.def (movx, partial_reg_dependency): Enable for
23639         m_SKYLAKE_AVX512.
23641 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
23643         PR inline-asm/84985
23644         * lra-constraints.c (process_alt_operands): Move setting
23645         this_alternative_matches below.
23647 2018-03-29  Martin Liska  <mliska@suse.cz>
23649         PR lto/84995.
23650         * doc/invoke.texi: Document how LTO works with debug info.
23651         Describe auto-load support of binutils.  Mention 'x86-64'
23652         as valid option value of -march option.
23654 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
23656         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
23658         PR c/85094
23659         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
23660         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
23661         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
23662         checking.
23664 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23666         PR target/84912
23667         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
23668         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
23669         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
23670         for RS6000_BTM_POWERPC64.
23671         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
23672         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
23673         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
23674         definition.
23675         (DIVDE): Use it.
23676         (DIVDEU): Likewise.
23678 2018-03-28  Carl Love  <cel@us.ibm.com>
23680         Revert
23681         2017-09-27  Carl Love  <cel@us.ibm.com>
23683         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23684         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23685         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23686         fctiw instruction.
23688 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23690         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23691         instead of __vector bool.
23692         (_mm_max_pu8): Likewise.
23693         (_mm_min_pi16): Likewise.
23695 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23697         PR target/84912
23698         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23699         (DIVWEUO): Likewise.
23700         (DIVDEO): Likewise.
23701         (DIVDEUO): Likewise.
23702         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23703         DIVWEUO and DIVDEUO.
23704         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23705         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23706         (div_extend): Likewise.
23707         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23708         builtin function.
23709         (__builtin_divweuo): Likewise.
23710         (__builtin_divdeo): Likewise.
23711         (__builtin_divdeuo): Likewise.
23713 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23715         PR target/85095
23716         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23717         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23719         PR tree-optimization/82004
23720         * gimple-match-head.c (optimize_pow_to_exp): New function.
23721         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23722         Don't fold to exp if optimize_pow_to_exp is false.
23724 2018-03-28  Martin Liska  <mliska@suse.cz>
23726         PR other/84819
23727         * calls.c (initialize_argument_information): Fix trailing space.
23728         * common.opt: Fix typo and provide better explanation for
23729         -fsanitize-coverage option.
23730         * config/i386/i386.opt: Fix typo.
23732 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23733             Martin Liska  <mliska@suse.cz>
23735         PR sanitizer/85081
23736         * gimplify.c (asan_poison_variable): Don't do the check for
23737         gimplify_omp_ctxp here.
23738         (gimplify_decl_expr): Do it here.
23739         (gimplify_target_expr): Likewise.
23741 2018-03-28  Martin Liska  <mliska@suse.cz>
23743         PR target/84988
23744         * config/i386/i386.c (ix86_function_arg_advance): Do not call
23745         chkp_type_bounds_count if MPX is not enabled.
23747 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
23749         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23751 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
23753         PR target/84914
23754         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23755         function to create the function decl for complex long double
23756         multiply and divide for -mabi=ieeelongdouble.
23757         (init_float128_ieee): Call it.
23759 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
23761         PR target/85044
23762         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23763         -fcf-protection=branch -mibt.
23764         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23766 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23768         PR target/81863
23769         * config/arm/arm.c (arm_valid_symbolic_address): Handle
23770         arm_word_relocations.
23772 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
23774         PR target/85056
23775         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23776         extern array declarations.
23778 2018-03-27  Richard Biener  <rguenther@suse.de>
23780         PR middle-end/84067
23781         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23782         explicit single_use checks.
23784 2018-03-27  Richard Biener  <rguenther@suse.de>
23786         PR tree-optimization/85082
23787         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23788         Valueize the VUSE.
23790 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23792         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23793         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23794         Turn on fasynchronous-unwind-tables and funwind-tables.
23796 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
23798         PR target/85073
23799         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23800         (*bmi_blsr_<mode>_ccz): Ditto.
23802 2018-03-26  Tom de Vries  <tom@codesourcery.com>
23804         PR tree-optimization/85063
23805         * omp-general.c (offloading_function_p): New function.  Factor out
23806         of ...
23807         * omp-offload.c (pass_omp_target_link::gate): ... here.
23808         * omp-general.h (offloading_function_p): Declare.
23809         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23810         with attribute omp declare target for offloading functions.
23812 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23814         PR tree-optimization/84005
23815         * tree-data-ref.h (get_base_for_alignment): Declare.
23816         * tree-data-ref.c (get_base_for_alignment_1): New function.
23817         (get_base_for_alignment): Likewise.
23818         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23819         get_base_for_alignment to find a suitable base object, instead
23820         of always using drb->base_address.
23822 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23824         PR inline-asm/85022
23825         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23826         known size by default.
23828 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23830         PR inline-asm/85030
23831         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23832         and non BLKmode operands.
23834 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23836         PR target/85026
23837         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23838         Clean up attributes.
23840 2018-03-23  Richard Biener  <rguenther@suse.de>
23842         PR debug/85020
23843         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23844         we are going to emit early debug for LTO.
23846 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23848         PR inline-asm/85034
23849         * function.c (match_asm_constraints_1): Don't optimize if input
23850         doesn't satisfy general_operand predicate for output's mode.
23852         PR inline-asm/85022
23853         * alias.c (write_dependence_p): Don't require for x_canonicalized
23854         non-VOIDmode if x has VOIDmode.
23856         PR sanitizer/85029
23857         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23858         just don't try to optimize it rather than assert it never happens.
23860 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23862         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23863         macro expansions for definition of ST_INTERNAL_<mode> and
23864         LD_INTERNAL_<mode> builtins.
23865         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23866         Remove prototype.
23867         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23868         function.
23869         (altivec_expand_st_builtin): Likewise.
23870         (altivec_expand_builtin): Remove calls to deleted functions.
23871         (rs6000_address_for_altivec): Delete this function.
23872         * config/rs6000/vector.md: Remove expands for
23873         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23875 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23877         PR target/84826
23878         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23879         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23880         re-computing once computed.
23881         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23882         (arm_init_machine_status): Initialize
23883         machine->static_chain_stack_bytes.
23885 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23887         PR target/84760
23888         * doc/extend.texi: Add four new prototypes for vec_ld.
23889         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23890         definitions for more logical presentation.
23891         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23892         entries for V1TI variants of __builtin_altivec_ld builtin.
23893         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23894         handling of V1TI variant of LVX icode pattern.
23895         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23896         (rs6000_gimple_fold_builtin): Likewise.
23897         (altivec_init_builtins): Add code to define
23898         __builtin_altivec_lvx_v1ti function.
23900 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23902         PR inline-asm/84941
23903         * function.c (match_asm_constraints_1): Don't do the optimization
23904         if input isn't a REG, SUBREG, MEM or constant.
23906 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23908         PR tree-optimization/84956
23909         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23910         bb_has_abnormal_pred.
23912 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23914         PR sanitizer/85018
23915         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23916         DECL_INITIAL (decl) to decl at the end.
23917         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23918         adjust the comment.
23920 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23922         * doc/extend.texi (__builtin_tgmath): Document when complex
23923         integer types are treated as _Complex _Float64.
23925 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23927         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23929 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23931         PR tree-optimization/84960
23932         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23933         if it is ENTRY block, move them into single succ of ENTRY in that case.
23935 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23937         PR tree-optimization/84811
23938         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23939         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23940         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23941         (known_subrange_p): Update accordingly.  Cast each value involved
23942         in the size comparison, rather than casting the result of the
23943         subtraction.
23945 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23947         PR tree-optimization/84982
23948         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23949         by flipping the least significant bit rather than all bits from
23950         bitpos to bitpos + bitsize - 1.
23952 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23954         * doc/extend.texi (Deprecated Features): Remove mention of
23955         long-deleted deprecations.
23957 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23959         PR jit/84288
23960         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23961         * configure: Regenerate.
23963 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23965         PR tree-optimization/83126
23966         * tree-parloops.c (num_phis): New function.
23967         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23969 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23971         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23972         mention anon-struct/union members and trailing attributes.
23974 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23976         PR tree-optimization/84969
23977         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23978         builtin memset partitions if they set different rhs values.
23980 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23982         PR rtl-optimization/84989
23983         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23984         VEC_DUPLICATE with scalar result mode.
23986 2018-03-21  Martin Liska  <mliska@suse.cz>
23988         PR ipa/84963
23989         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23990         not intended return statement.
23992 2018-03-21  Martin Liska  <mliska@suse.cz>
23994         PR target/84988
23995         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23996         (chkp_find_bound_slots_1): Limit number of iterations.
23998 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
24000         PR target/84838
24001         * Minor grammar fixes for x86 options.
24003 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24005         PR debug/84875
24006         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
24007         holding REG_CFA_RESTORE notes, instead turn them into a USE.
24009 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
24011         PR target/83789
24012         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
24013         (altivec_lvx_<mode>_1op): Likewise.
24014         (altivec_stvx_<mode>_2op): Likewise.
24015         (altivec_stvx_<mode>_1op): Likewise.
24016         (altivec_lvx_<VM2:mode>): New define_expand.
24017         (altivec_stvx_<VM2:mode>): Likewise.
24018         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
24019         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24020         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
24021         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24022         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
24023         (rs6000_gen_lvx): Likewise.
24024         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
24025         (altivec_expand_stv_builtin): Likewise.
24026         (altivec_expand_builtin): Likewise.
24027         * config/rs6000/vector.md: Likewise.
24029 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24031         PR target/82518
24032         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
24033         BYTES_BIG_ENDIAN.
24035 2018-03-20  Richard Biener  <rguenther@suse.de>
24037         PR target/84986
24038         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
24039         sign-conversions as zero, fall back to standard scalar_stmt
24040         cost for the rest.
24042 2018-03-20  Martin Liska  <mliska@suse.cz>
24044         PR ipa/84825
24045         * predict.c (rebuild_frequencies): Handle case when we have
24046         PROFILE_ABSENT, but flag_guess_branch_prob is false.
24048 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24050         PR target/84990
24051         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
24052         flag_section_anchors.
24053         * varasm.c (use_blocks_for_decl_p): Remove hack for
24054         dw2_force_const_mem.
24056         PR target/84845
24057         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
24058         to ...
24059         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
24060         be created, use lowpart_subreg of operands[0] rather than operands[0]
24061         itself.
24062         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
24063         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
24064         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
24065         and n constraint instead of aarch64_shift_imm_di and Usd.
24066         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
24067         (*aarch64_<optab>_reg_minus<mode>3): ... this.
24069 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
24071         PR target/82989
24072         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
24073         to favor GPR over NEON registers.
24074         (<shift>di3_neon): Likewise.
24076 2018-03-20  Tom de Vries  <tom@codesourcery.com>
24078         PR target/84952
24079         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
24080         (nvptx_process_pars): Emit bar.sync asap and alap.
24082 2018-03-20  Tom de Vries  <tom@codesourcery.com>
24084         PR target/84954
24085         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
24086         seen_label if seen_label is already set.
24088 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24090         PR target/84945
24091         * config/i386/i386.c (fold_builtin_cpu): For features above 31
24092         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
24093         Use 1U instead of 1.  Formatting fixes.
24095         PR c/84953
24096         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
24097         instead of TREE_TYPE (s1) for the return value.
24099 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
24101         PR tree-optimization/84946
24102         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
24103         bitsize + bitsize in poly_uint64 rather than poly_int64.
24105         PR sanitizer/78651
24106         * dwarf2asm.c: Include fold-const.c.
24107         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
24108         of decl rather than decl itself.
24110         PR rtl-optimization/84643
24111         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
24113 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
24115         PR sanitizer/78651
24116         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
24117         calling assemble_variable.
24119 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
24121         PR target/81647
24122         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
24123         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
24125 2018-03-19  Jim Wilson  <jimw@sifive.com>
24127         PR bootstrap/84856
24128         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
24129         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
24130         (riscv_first_stack_step): Likewise.
24131         (riscv_option_override): Use STACK_BOUNDARY instead of
24132         MIN_STACK_BOUNDARY.
24133         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
24134         MIN_STACK_BOUNDARY.
24135         (BIGGEST_ALIGNMENT): Set to 128.
24136         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
24137         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
24138         STACK_BOUNDARY.
24140 2018-03-19  Richard Biener  <rguenther@suse.de>
24142         PR tree-optimization/84933
24143         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
24144         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
24146 2018-03-19  Richard Biener  <rguenther@suse.de>
24148         PR tree-optimization/84859
24149         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
24150         (cond_if_else_store_replacement): Perform sinking operation on
24151         single-store BBs regardless of MAX_STORES_TO_SINK setting.
24152         Generalize what a BB with a single eligible store is.
24154 2018-03-19  Richard Biener  <rguenther@suse.de>
24156         PR tree-optimization/84929
24157         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
24158         chrec_is_positive against non-chrec arg.
24160 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
24162         PR target/84711
24163         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
24165 2018-03-18  Martin Liska  <mliska@suse.cz>
24167         PR rtl-optimization/84635
24168         * regrename.c (build_def_use): Use matches_mode only when
24169         matches >= 0.
24171 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
24173         PR tree-optimization/84913
24174         * tree-vect-loop.c (vectorizable_reduction): Don't try to
24175         vectorize chains of COND_EXPRs.
24177 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24179         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
24181 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24183         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
24185 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24187         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
24189 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
24190             Kito Cheng  <kito.cheng@gmail.com>
24192         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
24193         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
24194         (nds32_adjust_reg_alloc_order): New function.
24195         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
24197 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
24199         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
24200         nds32_print_operand, nds32_print_operand_address): Use
24201         HOST_WIDE_INT_PRINT_DEC instead.
24203 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
24205         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
24207 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
24209         PR target/84902
24210         * config/i386/i386.c (initial_ix86_tune_features,
24211         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
24212         unsigned long long.
24213         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
24214         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
24215         rather than 1u << ix86_tune.  Formatting fix.
24216         (ix86_option_override_internal): Change ix86_arch_mask from
24217         unsigned int to unsigned HOST_WIDE_INT, initialize to
24218         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
24219         (ix86_function_specific_restore): Likewise.
24221 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24223         PR target/84899
24224         * postreload.c (reload_combine_recognize_pattern): Perform
24225         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
24226         truncate_int_for_mode the result for the destination's mode.
24228         PR c/84909
24229         * hsa-gen.c (mem_type_for_type): Fix comment typo.
24230         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
24231         Likewise.
24232         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24233         Likewise.
24235 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
24237         PR target/84876
24238         * lra-assigns.c (lra_split_hard_reg_for): Don't use
24239         regno_allocno_class_array and sorted_pseudos.
24240         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
24241         insns where regno is used.
24243 2018-03-16  Martin Liska  <mliska@suse.cz>
24245         PR ipa/84833
24246         * multiple_target.c (create_dispatcher_calls): Redirect
24247         reference in the symbol table.
24249 2018-03-16  Martin Liska  <mliska@suse.cz>
24251         PR ipa/84722
24252         * multiple_target.c (create_dispatcher_calls): Redirect also
24253         an alias.
24255 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24257         PR c++/79937
24258         PR c++/82410
24259         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
24260         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
24261         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
24263 2018-03-16  Julia Koval  <julia.koval@intel.com>
24265         * doc/invoke.texi (Skylake Server): Add CLWB.
24266         Cannonlake): Remove CLWB.
24268 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24270         PR tree-optimization/84841
24271         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
24272         1 << 3.
24273         (FLOAT_ONE_CONST_TYPE): Define.
24274         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
24275         (sort_by_operand_rank): Put entries with higher constant_type last
24276         rather than first to match comments.
24278 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
24280         * config/nios2/nios2.md (movsi_internal): Fix thinko in
24281         split predicate.
24283 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
24285         PR c++/79085
24286         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
24287         check and use address of target always.
24289 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
24291         PR target/84574
24292         * config/i386/i386.c (indirect_thunk_needed): Update comments.
24293         (indirect_thunk_bnd_needed): Likewise.
24294         (indirect_thunks_used): Likewise.
24295         (indirect_thunks_bnd_used): Likewise.
24296         (indirect_return_needed): New.
24297         (indirect_return_bnd_needed): Likewise.
24298         (output_indirect_thunk_function): Add a bool argument for
24299         function return.
24300         (output_indirect_thunk_function): Don't generate alias for
24301         function return thunk.
24302         (ix86_code_end): Call output_indirect_thunk_function to generate
24303         function return thunks.
24304         (ix86_output_function_return): Set indirect_return_bnd_needed
24305         and indirect_return_needed instead of indirect_thunk_bnd_needed
24306         and indirect_thunk_needed.
24308 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
24310         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
24311         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
24312         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
24314 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
24315             Paul Hua <paul.hua.gm@gmail.com>
24317         PR c/84852
24318         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
24320 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
24322         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
24323         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
24324         resp. SFmode cases.
24326 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
24328         PR target/84711
24329         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
24330         instead of GET_MODE_SIZE when comparing Units.
24332 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
24334         PR target/68256
24335         * varasm.c (hash_section): Return an unchangeble hash value
24336         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
24337         Return !aarch64_can_use_per_function_literal_pools_p ().
24339 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
24341         PR target/84860
24342         * optabs.c (emit_conditional_move): Pass address of cmode's copy
24343         rather than address of cmode as last argument to prepare_cmp_insn.
24345 2018-03-15  Julia Koval  <julia.koval@intel.com>
24347         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
24348         F_AVX512VNNI, F_AVX512BITALG): New.
24350 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
24352         PR target/83451
24353         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
24354         insn for floating-point loads and stores.
24356 2018-03-14  Carl Love  <cel@us.ibm.com>
24358         * config/rs6000/rs6000-c.c: Add macro definitions for
24359         ALTIVEC_BUILTIN_VEC_PERMXOR.
24360         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
24361         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
24362         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
24363         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
24364         UNSPEC_VPERMXOR.
24365         * config/doc/extend.texi: Add prototypes for vec_permxor.
24367 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
24369         PR c/84852
24370         * diagnostic-show-locus.c (class layout_point): Convert m_line
24371         from int to linenum_type.
24372         (line_span::comparator): Use linenum "compare" function when
24373         comparing line numbers.
24374         (test_line_span): New function.
24375         (layout_range::contains_point): Convert param "row" from int to
24376         linenum_type.
24377         (layout_range::intersects_line_p): Likewise.
24378         (layout::will_show_line_p): Likewise.
24379         (layout::print_source_line): Likewise.
24380         (layout::should_print_annotation_line_p): Likewise.
24381         (layout::print_annotation_line): Likewise.
24382         (layout::print_leading_fixits): Likewise.
24383         (layout::annotation_line_showed_range_p): Likewise.
24384         (struct line_corrections): Likewise for field m_row.
24385         (line_corrections::line_corrections): Likewise for param "row".
24386         (layout::print_trailing_fixits): Likewise.
24387         (layout::get_state_at_point): Likewise.
24388         (layout::get_x_bound_for_row): Likewise.
24389         (layout::print_line): Likewise.
24390         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
24391         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
24392         * input.c (selftest::test_linenum_comparisons): New function.
24393         (selftest::input_c_tests): Call it.
24394         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
24395         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
24396         * selftest.h (ASSERT_GT): New macro.
24397         (ASSERT_GT_AT): New macro.
24398         (ASSERT_LT): New macro.
24399         (ASSERT_LT_AT): New macro.
24401 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
24403         PR rtl-optimization/84780
24404         * combine.c (distribute_links): Don't make a link based on pc_rtx.
24406 2018-03-14  Martin Liska  <mliska@suse.cz>
24408         * tree.c (record_node_allocation_statistics): Use
24409         get_stats_node_kind.
24410         (get_stats_node_kind): New function extracted from
24411         record_node_allocation_statistics.
24412         (free_node): Use get_stats_node_kind.
24414 2018-03-14  Richard Biener  <rguenther@suse.de>
24416         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
24417         that the value-set of ANTIC_IN doesn't grow.
24419         Revert
24420         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24421         member.
24422         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24423         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24425 2018-03-14  Julia Koval  <julia.koval@intel.com>
24427         * config.gcc (icelake-client, icelake-server): New.
24428         (icelake): Remove.
24429         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
24430         (initial_ix86_arch_features): Ditto.
24431         (PTA_SKYLAKE): Add SGX.
24432         (PTA_ICELAKE): Remove.
24433         (PTA_ICELAKE_CLIENT): New.
24434         (PTA_ICELAKE_SERVER): New.
24435         (ix86_option_override_internal): Split up icelake on icelake client and
24436         icelake server.
24437         (get_builtin_code_for_version): Ditto.
24438         (fold_builtin_cpu): Ditto.
24439         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
24440         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
24441         * config/i386/i386.h (processor_type): Ditto.
24442         * doc/invoke.texi: Ditto.
24444 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
24446         PR sanitizer/83392
24447         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
24448         INTEGER_CST offset, add it together with bitpos / 8 and
24449         sign extend based on POINTER_SIZE.
24451         PR target/84844
24452         Revert
24453         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
24455         PR target/78090
24456         * config/i386/constraints.md (Yc): New register constraint.
24457         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
24458         Use Yc constraint for alternative 2 of operand 0.  Remove
24459         preferred_for_speed attribute.
24461 2018-03-14  Richard Biener  <rguenther@suse.de>
24463         PR tree-optimization/84830
24464         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
24465         with the old one to avoid oscillations.
24467 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
24469         PR target/83712
24470         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
24471         pseudos.
24472         (assign_by_spills): Return a flag of reload assignment failure.
24473         Do not process the reload assignment failures.  Do not spill other
24474         reload pseudos if they has the same reg class.  Update n if
24475         necessary.
24476         (lra_assign): Add a return arg.  Set up from the result of
24477         assign_by_spills call.
24478         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24479         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24480         usage_insns if it is not NULL.
24481         (spill_hard_reg_in_range): New function.
24482         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24483         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24484         function prototypes.
24485         (lra_assign): Change prototype.
24486         * lra.c (lra): Add code to deal with fails by splitting hard reg
24487         live ranges.
24489 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
24491         * config/riscv/riscv.opt (mrelax): New option.
24492         * config/riscv/riscv.c (riscv_file_start): Emit ".option
24493         "norelax" when riscv_mrelax is disabled.
24494         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
24496 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24498         PR target/84743
24499         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
24500         reassociation for int modes.
24502 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
24504         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
24505         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
24506         for big-endian.
24507         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
24508         * config/aarch64/aarch64-sve.md
24509         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
24510         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
24511         (*extend<mode><Vwide>2): Rename to...
24512         (aarch64_sve_extend<mode><Vwide>2): ...this.
24513         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
24514         renaming the old pattern to...
24515         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
24516         unsigned packs.
24517         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
24518         define_expand, renaming the old pattern to...
24519         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
24520         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
24521         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
24522         account when deciding which SVE instruction the optab should use.
24523         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
24525 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
24527         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
24528         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
24529         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
24530         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
24531         (tlsdesc_small_<mode>): Turn a define_expand and use
24532         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
24533         (tlsdesc_small_advsimd_<mode>): ...this.
24534         (tlsdesc_small_sve_<mode>): New pattern.
24536 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
24538         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
24539         (UNSPEC_UMUL_HIGHPART): New constants.
24540         (MUL_HIGHPART): New int iteraor.
24541         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
24542         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
24543         define_expand.
24544         (*<su>mul<mode>3_highpart): New define_insn.
24546 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
24548         PR lto/84805
24549         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
24550         incomplete types.
24552 2018-03-13  Martin Liska  <mliska@suse.cz>
24554         PR ipa/84658.
24555         * (sem_item_optimizer::sem_item_optimizer): Initialize new
24556         vector.
24557         (sem_item_optimizer::~sem_item_optimizer): Release it.
24558         (sem_item_optimizer::merge_classes): Register variable aliases.
24559         (sem_item_optimizer::fixup_pt_set): New function.
24560         (sem_item_optimizer::fixup_points_to_sets): Likewise.
24561         * ipa-icf.h: Declare new variables and functions.
24563 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
24565         PR middle-end/84834
24566         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
24567         integer_pow2p@2 and test integer_pow2p in condition.
24568         (A < 0 ? C : 0): Similarly for @1.
24570         PR middle-end/84831
24571         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
24572         characters starting at p contain '\0' character, don't look beyond
24573         that.
24575         PR target/84827
24576         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
24577         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
24579         PR target/84828
24580         * reg-stack.c (change_stack): Change update_end var from int to
24581         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
24582         also call set_block_for_insn on the newly added insns and rescan.
24584         PR target/84786
24585         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
24586         on the last operand.
24588         PR c++/84704
24589         * tree.c (stabilize_reference_1): Return save_expr (e) for
24590         STATEMENT_LIST even if it doesn't have side-effects.
24592 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
24594         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
24596 2018-03-12  Renlin Li  <renlin.li@arm.com>
24598         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
24599         aarch64_output_scalar_simd_mov_immediate.
24601 2018-03-12  Martin Sebor  <msebor@redhat.com>
24603         PR tree-optimization/83456
24604         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
24605         for perfectly overlapping calls to memcpy.
24606         (gimple_fold_builtin_memory_chk): Same.
24607         (gimple_fold_builtin_strcpy): Handle no-warning.
24608         (gimple_fold_builtin_stxcpy_chk): Same.
24609         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
24611 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
24613         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
24614         parameter.  Use it for SFmode.
24615         (rs6000_function_arg_advance_1): Adjust.
24616         (rs6000_function_arg): Adjust.
24617         (rs6000_gimplify_va_arg): Pass false for that new parameter.
24619 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
24621         PR rtl-optimization/84169
24622         PR rtl-optimization/84780
24623         * combine.c (can_combine_p): Check for a 2-insn combination whether
24624         the destination register is used between the two insns, too.
24626 2018-03-12  Richard Biener  <rguenther@suse.de>
24628         PR tree-optimization/84803
24629         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
24630         for refs DR analysis didn't process.
24632 2018-03-12  Richard Biener  <rguenther@suse.de>
24634         PR tree-optimization/84777
24635         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
24636         force-vectorize loops ignore whether we are optimizing for size.
24638 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
24640         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
24641         (TARGET_MD_ASM_ADJUST): Define.
24643 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
24644             Kito Cheng  <kito.cheng@gmail.com>
24645             Chung-Ju Wu  <jasonwucj@gmail.com>
24647         * config/nds32/nds32.c (nds32_compute_stack_frame,
24648         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
24649         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
24650         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
24651         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
24652         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
24653         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
24654         * config/nds32/nds32.md (prologue, epilogue): Use macro
24655         NDS32_V3PUSH_AVAILABLE_P to do checking.
24657 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
24659         PR debug/58150
24660         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
24661         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
24662         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
24663         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
24664         addition of most attributes on !orig_type_die or the attribute not
24665         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
24667 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24668             Chung-Ju Wu  <jasonwucj@gmail.com>
24670         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
24671         __NDS32_VH__ macro.
24672         * config/nds32/nds32.opt (mvh): New option.
24674 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24675             Chung-Ju Wu  <jasonwucj@gmail.com>
24677         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
24678         function.
24679         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
24680         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
24681         definition.
24683 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24684             Chung-Ju Wu  <jasonwucj@gmail.com>
24686         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24687         function.
24688         * config/nds32/nds32-multiple.md (strlensi): New pattern.
24689         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24691 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
24692             Kito Cheng  <kito.cheng@gmail.com>
24693             Chung-Ju Wu  <jasonwucj@gmail.com>
24695         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24696         UNSPEC_FFMISM and UNSPEC_FLMISM.
24697         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24698         for ffb, ffmism and flmism.
24699         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24700         (unspec_ffmism): Ditto.
24701         (unspec_flmism): Ditto.
24702         (nds32_expand_builtin_impl): Check if string extension is available.
24703         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24704         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24706 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
24708         Reverting patch:
24709         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24711         PR target/83712
24712         * lra-assigns.c (assign_by_spills): Return a flag of reload
24713         assignment failure.  Do not process the reload assignment
24714         failures.  Do not spill other reload pseudos if they has the same
24715         reg class.
24716         (lra_assign): Add a return arg.  Set up from the result of
24717         assign_by_spills call.
24718         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24719         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24720         usage_insns if it is not NULL.
24721         (spill_hard_reg_in_range): New function.
24722         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24723         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24724         function prototypes.
24725         (lra_assign): Change prototype.
24726         * lra.c (lra): Add code to deal with fails by splitting hard reg
24727         live ranges.
24729 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
24731         PR target/84807
24732         * config/i386/i386.opt: Replace Enforcment with Enforcement.
24734 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
24736         PR debug/84620
24737         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24738         (dw_val_node): Add val_symbolic_view.
24739         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24740         (symview_upper_bound): New.
24741         (new_line_info_table): Initialize symviews_since_reset.
24742         (dwarf2out_source_line): Count symviews_since_reset and set
24743         symview_upper_bound.
24744         (dw_val_equal_p): Handle symview.
24745         (add_AT_symview): New.
24746         (print_dw_val): Handle symview.
24747         (attr_checksum, attr_checksum_ordered): Likewise.
24748         (same_dw_val_p, size_of_die): Likewise.
24749         (value_format, output_die): Likewise.
24750         (add_high_low_attributes): Use add_AT_symview for entry_view.
24751         (dwarf2out_finish): Reset symview_upper_bound, clear
24752         zero_view_p.
24754 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
24756         PR target/83969
24757         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24758         Add strict argument and use it.
24759         (rs6000_split_multireg_move): Update for new strict argument.
24760         (mem_operand_gpr): Disallow all non-offsettable addresses.
24761         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24763 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
24765         PR target/84772
24766         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24767         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24768         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24770         PR c++/84767
24771         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24772         decl, use remap_type if we want to use the type.
24774 2018-03-09  Martin Sebor  <msebor@redhat.com>
24776         PR tree-optimization/84526
24777         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24778         Remove dead code.
24779         (builtin_access::generic_overlap): Be prepared to handle non-array
24780         base objects.
24782 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
24784         PR rtl-optimization/84682
24785         * lra-constraints.c (process_address_1): Check is_address flag
24786         for address constraints.
24787         (process_alt_operands): Likewise.
24788         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24789         preprocess_constraints.
24790         * recog.h (preprocess_constraints): Add oploc parameter.
24791         Adjust callers.
24792         * recog.c (preprocess_constraints): Test address_operand for
24793         CT_ADDRESS constraints.
24795 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24797         PR target/83712
24798         * lra-assigns.c (assign_by_spills): Return a flag of reload
24799         assignment failure.  Do not process the reload assignment
24800         failures.  Do not spill other reload pseudos if they has the same
24801         reg class.
24802         (lra_assign): Add a return arg.  Set up from the result of
24803         assign_by_spills call.
24804         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24805         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24806         usage_insns if it is not NULL.
24807         (spill_hard_reg_in_range): New function.
24808         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24809         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24810         function prototypes.
24811         (lra_assign): Change prototype.
24812         * lra.c (lra): Add code to deal with fails by splitting hard reg
24813         live ranges.
24815 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24817         PR target/83193
24818         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24819         Accept complain bool parameter.  Only emit errors if it is true.
24820         (arm_parse_cpu_option_name): Likewise.
24821         (arm_target_thumb_only): Adjust callers of the above.
24822         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24823         prototype to take a default true bool parameter.
24824         (arm_parse_arch_option_name): Likewise.
24826 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24827             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24829         PR jit/64089
24830         PR jit/84288
24831         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24832         * configure: Regenerate.
24833         * configure.ac ("linker --version-script option"): New.
24834         ("linker soname option"): New.
24836 2018-03-09  Richard Biener  <rguenther@suse.de>
24838         PR tree-optimization/84775
24839         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24840         immediate uses of predicate stmts and mark them modified.
24842         Revert
24843         PR tree-optimization/84178
24844         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24845         to caller.
24846         (version_loop_for_if_conversion): Delay update_ssa call.
24847         (tree_if_conversion): Delay update_ssa until after predicate
24848         insertion.
24850 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24852         PR target/84763
24853         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24854         when the function accesses prior frames.
24856 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24858         PR debug/84456
24859         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24860         gen_llsym, otherwise call maybe_gen_llsym.
24862         PR inline-asm/84742
24863         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24864         has ',' character inside of it.
24866 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24868         PR target/84748
24869         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24870         as clobbering CC_REGNUM.
24872 2018-03-08  Richard Biener  <rguenther@suse.de>
24874         PR middle-end/84552
24875         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24876         (follow_copies_to_constant): Do not follow SSA names registered
24877         for update.
24879 2018-03-08  Richard Biener  <rguenther@suse.de>
24881         PR tree-optimization/84178
24882         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24883         to caller.
24884         (version_loop_for_if_conversion): Delay update_ssa call.
24885         (tree_if_conversion): Delay update_ssa until after predicate
24886         insertion.
24888 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24890         PR tree-optimization/84178
24891         * tree-if-conv.c (release_bb_predicate): Remove the
24892         the assertion that the stmts have NULL use_ops.
24893         Discard the statements, asserting that they haven't
24894         yet been added to a BB.
24896 2018-03-08  Richard Biener  <rguenther@suse.de>
24898         PR tree-optimization/84746
24899         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24900         (phi_translate): Pass in destination ANTIC_OUT set.
24901         (phi_translate_1): Likewise.  For a simplified result lookup
24902         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24903         (phi_translate_set): Adjust.
24904         (do_pre_regular_insertion): Likewise.
24905         (do_pre_partial_partial_insertion): Likewise.
24907 2018-03-08  Martin Liska  <mliska@suse.cz>
24909         PR gcov-profile/84735
24910         * doc/gcov.texi: Document usage of profile files.
24911         * gcov-io.h: Document changes in the format.
24913 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24915         PR debug/84404
24916         PR debug/84408
24917         * dwarf2out.c (struct dw_line_info_table): Update comments for
24918         view == -1.
24919         (FORCE_RESET_NEXT_VIEW): New.
24920         (FORCE_RESETTING_VIEW_P): New.
24921         (RESETTING_VIEW_P): Check for -1 too.
24922         (ZERO_VIEW_P): Likewise.
24923         (new_line_info_table): Force-reset next view.
24924         (dwarf2out_begin_function): Likewise.
24925         (dwarf2out_source_line): Simplify zero_view_p initialization.
24926         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24927         view directly.  Omit view when omitting .loc at line 0.
24929 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24931         PR tree-optimization/84740
24932         * tree-switch-conversion.c (process_switch): Call build_constructors
24933         only if info.phi_count is non-zero.
24935         PR tree-optimization/84739
24936         * tree-tailcall.c (find_tail_calls): Check call arguments against
24937         DECL_ARGUMENTS (current_function_decl) rather than
24938         DECL_ARGUMENTS (func) when checking for tail recursion.
24940 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24942         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24943         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24944         Volker Reichelt's entry and add entries for people that perform
24945         GCC fuzzy testing and report numerous bugs.
24947 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24949         PR target/82411
24950         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24951         readonly data in sdata, if that is disabled.
24952         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24953         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24954         -mreadonly-in-sdata option.
24956 2018-03-07  Martin Sebor  <msebor@redhat.com>
24958         PR tree-optimization/84468
24959         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24960         basic block when looking for nul assignment.
24962 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24964         PR target/84277
24965         * except.h (output_function_exception_table): Adjust prototype.
24966         * except.c (output_function_exception_table): Remove FNNAME parameter
24967         and add SECTION parameter.  Ouput one part of the table at a time.
24968         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24969         the first part of the exception table and emit unwind directives.
24970         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24971         (i386_pe_seh_cold_init): Likewise.
24972         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24973         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24974         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24975         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24976         * config/i386/winnt.c: Include except.h.
24977         (struct seh_frame_state): Add reg_offset, after_prologue and
24978         in_cold_section fields.
24979         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24980         (i386_pe_seh_cold_init): New function.
24981         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24982         to seh->in_cold_section.
24983         (seh_emit_push): Record the offset of the push.
24984         (seh_emit_save): Record the offet of the save.
24985         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24986         Test seh->after_prologue to disregard the epilogue.
24987         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24988         (i386_pe_end_cold_function): New function.
24990 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24992         PR fortran/84565
24993         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24994         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24996         PR c++/84704
24997         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24998         on tmp_var.
24999         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
25000         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
25002         PR middle-end/84723
25003         * multiple_target.c: Include tree-inline.h and intl.h.
25004         (expand_target_clones): Diagnose and fail if node->definition and
25005         !tree_versionable_function_p (node->decl).
25007 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
25009         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
25010         sprint_ul.
25011         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
25012         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
25013         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
25015 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
25017         PR target/84710
25018         * combine.c (try_combine): Use reg_or_subregno instead of handling
25019         just paradoxical SUBREGs and REGs.
25021 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
25023         * config/arc/arc.c (arc_finalize_pic): Remove function.
25024         (arc_must_save_register): We use single base PIC register, remove
25025         checks to save/restore the PIC register.
25026         (arc_expand_prologue): Likewise.
25027         * config/arc/arc-protos.h (arc_set_default_type_attributes):
25028         Remove.
25029         (arc_verify_short): Likewise.
25030         (arc_attr_type): Likewise.
25031         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
25032         (walk_stores): Likewise.
25033         (arc_address_cost): Make it static.
25034         (arc_verify_short): Likewise.
25035         (branch_dest): Likewise.
25036         (arc_attr_type): Likewise.
25037         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
25038         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
25039         (arc_final_prescan_insn): Remove inserting the nops due to
25040         hardware hazards.  It is done in reorg step.
25041         (insn_length_variant_t): Remove.
25042         (insn_length_parameters_t): Likewise.
25043         (arc_insn_length_parameters): Likewise.
25044         (arc_get_insn_variants): Likewise.
25045         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
25047 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
25049         PR inline-asm/84683
25050         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
25051         assertion failure.
25053         PR tree-optimization/84687
25054         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
25055         on new_node->decl.
25056         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
25058 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25060         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
25061         Rename to ppc_speculation_barrier.
25062         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
25063         __builtin_ppc_speculation_barrier.
25065 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
25067         PR target/84700
25068         * combine.c (combine_simplify_rtx): Don't try to simplify if
25069         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
25070         are equal to x.
25072 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
25074         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
25075         to 32 bytes when compiling for POWER9.
25077 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
25079         PR target/84564
25080         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
25081         regparm >= 3 with no arg reg available also for calls with
25082         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
25084         PR target/84524
25085         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
25086         orig,vex.
25087         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
25089 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
25091         PR target/84264
25092         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
25094 2018-03-05  Richard Biener  <rguenther@suse.de>
25096         PR tree-optimization/84486
25097         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
25098         When inserting a __builtin_assume_aligned call set the LHS
25099         SSA name alignment info accordingly.
25101 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
25103         PR tree-optimization/84114
25104         * config/aarch64/aarch64.c (aarch64_reassociation_width)
25105         Avoid reassociation of FLOAT_MODE addition.
25107 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
25109         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
25110         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
25111         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
25112         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
25113         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
25114         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
25115         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
25116         and -mwbnoinvd.
25117         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
25118         __builtin_ia32_wbinvd): New builtins.
25119         (SPECIAL_ARGS2): New.
25120         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
25121         (SPECIAL_ARGS2): New.
25122         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
25123         (ix86_valid_target_attribute_inner_p): Ditto.
25124         (ix86_init_mmx_sse_builtins): Add special_args2.
25125         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
25126         TARGET_WBNOINVD_P): New.
25127         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
25128         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
25129         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
25130         * config/i386/immintrin.h (_wbinvd): New intrinsic.
25131         * config/i386/pconfigintrin.h: New file.
25132         * config/i386/wbnoinvdintrin.h: Ditto.
25133         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
25134         wbnoinvdintrin.h.
25135         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
25137 2018-03-05  Richard Biener  <rguenther@suse.de>
25139         PR tree-optimization/84670
25140         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
25141         member.
25142         (BB_VISITED_WITH_VISITED_SUCCS): New define.
25143         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
25144         (compute_antic_aux): Only assert the number of values in ANTIC_IN
25145         doesn't grow if all successors (recursively) were visited at least
25146         once.
25148 2018-03-05  Richard Biener  <rguenther@suse.de>
25150         PR tree-optimization/84650
25151         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
25152         if executed in the loop pipeline.
25154 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
25156         * doc/configfiles.texi (Configuration Files): Move info about
25157         conditionalizing $target-protos.h to...
25158         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
25159         differs from $target-protos.h.
25161 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
25162             Chung-Ju Wu  <jasonwucj@gmail.com>
25164         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
25165         * config/nds32/nds32-multiple.md (setmemsi): Define.
25166         * config/nds32/nds32-memory-manipulation.c
25167         (nds32_gen_dup_4_byte_to_word_value): New.
25168         (emit_setmem_word_loop): New.
25169         (emit_setmem_byte_loop): New.
25170         (nds32_expand_setmem_loop): New.
25171         (nds32_expand_setmem_loop_v3m): New.
25172         (nds32_expand_setmem_unroll): New.
25173         (nds32_expand_setmem): New.
25175 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25176             Chung-Ju Wu  <jasonwucj@gmail.com>
25178         * config/nds32/nds32-memory-manipulation.c
25179         (nds32_emit_load_store): New.
25180         (nds32_emit_post_inc_load_store): New.
25181         (nds32_emit_mem_move): New.
25182         (nds32_emit_mem_move_block): New.
25183         (nds32_expand_movmemsi_loop_unknown_size): New.
25184         (nds32_expand_movmemsi_loop_known_size): New.
25185         (nds32_expand_movmemsi_loop): New.
25186         (nds32_expand_movmemsi_unroll): New.
25187         (nds32_expand_movmemqi): Rename ...
25188         (nds32_expand_movmemsi): ... to this.
25189         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
25190         (movmemsi): ... to this.
25191         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
25192         (nds32_expand_movmemsi): ... to this.
25194 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25195             Monk Chiang  <sh.chiang04@gmail.com>
25196             Chung-Ju Wu  <jasonwucj@gmail.com>
25198         * config/nds32/nds32-protos.h
25199         (nds32_expand_load_multiple): New arguments.
25200         (nds32_expand_store_multiple): Ditto.
25201         (nds32_valid_multiple_load_store): Rename ...
25202         (nds32_valid_multiple_load_store_p): ... to this.
25203         * config/nds32/nds32-memory-manipulation.c
25204         (nds32_expand_load_multiple): Refine implementation.
25205         (nds32_expand_store_multiple): Ditto.
25206         * config/nds32/nds32-multiple.md
25207         (load_multiple): Update nds32_expand_load_multiple interface.
25208         (store_multiple): Update nds32_expand_store_multiple interface.
25209         * config/nds32/nds32-predicates.c
25210         (nds32_valid_multiple_load_store): Rename ...
25211         (nds32_valid_multiple_load_store_p): ... to this and refine
25212         implementation.
25213         * config/nds32/predicates.md
25214         (nds32_load_multiple_and_update_address_operation): New predicate.
25215         (nds32_store_multiple_and_update_address_operation): New predicate.
25217 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25218             Chung-Ju Wu  <jasonwucj@gmail.com>
25220         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
25221         (combo): New attribute.
25222         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
25224 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
25226         * config/nds32/nds32.opt: Change -mcmodel= default value.
25228 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
25229             Monk Chiang  <sh.chiang04@gmail.com>
25230             Chung-Ju Wu  <jasonwucj@gmail.com>
25232         * config/nds32/constants.md (unspec_element): New enum.
25233         * config/nds32/constraints.md (Umw): New constraint.
25234         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
25235         * config/nds32/nds32-intrinsic.md: Likewise.
25236         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
25237         (nds32_valid_smw_lwm_base_p): New.
25238         (nds32_output_smw_single_word): New.
25239         (nds32_output_lmw_single_word): New.
25240         (nds32_expand_unaligned_load): New.
25241         (nds32_expand_unaligned_store): New.
25242         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
25243         (nds32_output_smw_single_word): Declare.
25244         (nds32_output_lmw_single_word): Declare.
25245         (nds32_expand_unaligned_load): Declare.
25246         (nds32_expand_unaligned_store): Declare.
25247         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
25248         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
25249         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
25250         NDS32_BUILTIN_UASTORE_DW.
25251         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
25252         predicate.
25254 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
25255             Kito Cheng  <kito.cheng@gmail.com>
25256             Chung-Ju Wu  <jasonwucj@gmail.com>
25258         * config/nds32/nds32-intrinsic.c
25259         (nds32_expand_builtin_null_ftype_reg): Delete.
25260         (nds32_expand_builtin_reg_ftype_imm): Ditto.
25261         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
25262         (nds32_read_argument): New.
25263         (nds32_legitimize_target): Ditto.
25264         (nds32_legitimize_argument): Ditto.
25265         (nds32_check_constant_argument): Ditto.
25266         (nds32_expand_unop_builtin): Ditto.
25267         (nds32_expand_unopimm_builtin): Ditto.
25268         (nds32_expand_binop_builtin): Ditto.
25269         (nds32_builtin_decl_impl): Ditto.
25270         (builtin_description): Ditto.
25271         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
25272         (nds32_init_builtins_impl): Ditto.
25273         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
25274         (nds32_builtin_decl): New.
25275         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
25276         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
25278 2018-03-02  Jeff Law  <law@redhat.com>
25280         * reorg.c (stop_search_p): Handle DEBUG_INSN.
25281         (redundant_insn, fill_simple_delay_slots): Likewise.
25282         (fill_slots_from_thread): Likewise.
25283         * resource.c (mark_referenced_resources): Likewise.
25284         (mark_set_resources, find_dead_or_set_registers): Likewise.
25286 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25288         * substring-locations.h (format_warning_va): Formatting fix for
25289         ATTRIBUTE_GCC_DIAG.
25290         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
25291         argument.
25292         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
25293         * substring-locations.c: Include intl.h.
25294         (format_warning_va): Turned into small wrapper around
25295         format_warning_n_va, renamed to ...
25296         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
25297         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
25298         use ngettext.
25299         (format_warning_at_substring_n): New function.
25300         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
25301         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
25302         format_warning_at_substring with just a shorter name instead of
25303         const function pointer.
25304         (fmtwarn_n): New function.
25305         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
25306         appropriate, get rid of all the fmtstr temporaries, move conditionals
25307         with G_() wrapped string literals directly into fmtwarn arguments,
25308         cast dir.len to (int), formatting fixes.
25310 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
25312         * doc/invoke.texi: Remove "Cilk Plus" references.
25314 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25315             Richard Biener  <rguenther@suse.de>
25317         PR ipa/84628
25318         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
25319         for error or warning attributes if CALL_FROM_THUNK_P is set.
25320         Formatting fixes.
25322 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25324         PR target/56540
25325         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
25326         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
25328         PR target/56540
25329         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
25330         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
25332         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
25333         instead of -1U in last predictors element's probability member.
25335 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
25337         PR ipa/83983
25338         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
25339         arguments if they are comparable.
25341 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
25343         PR tree-optimization/84634
25344         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
25345         masks and masked_loop_p with a single loop_masks, making sure it's
25346         null for bb vectorization.
25348 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
25350         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
25351         (vect_analyze_data_ref_access): Use loop->safe_len rather than
25352         loop->force_vectorize to check whether there is no alias.
25354 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25356         PR target/84614
25357         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
25358         prototypes.
25359         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
25360         comments.
25361         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
25362         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
25363         instead of a loop around prev_real_insn.
25364         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
25365         prev_real_insn.
25367         PR inline-asm/84625
25368         * config/i386/i386.c (ix86_print_operand): Use conditional
25369         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
25370         zero vector.
25372 2018-03-02  Richard Biener  <rguenther@suse.de>
25374         PR tree-optimization/84427
25375         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
25376         (bitmap_set_subtract_values): Rewrite to handle multiple
25377         exprs per value.
25378         (clean): Likewise.
25379         (prune_clobbered_mems): Likewise.
25380         (phi_translate): Take edge instead of pred/phiblock.
25381         (phi_translate_1): Likewise.
25382         (phi_translate_set): Likewise.  Insert all translated
25383         exprs for a value into the set, keeping possibly multiple
25384         expressions per value.
25385         (compute_antic_aux): Adjust for phi_translate changes.
25386         When intersecting union the expressions and prune those
25387         not in the final value set, keeping possibly multiple
25388         expressions per value.  Do not use value-insertion
25389         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
25390         all expressions.  Add verification that the value-sets
25391         only shrink during iteration.
25392         (compute_partial_antic_aux): Adjust for the phi_translate changes.
25393         (do_pre_regular_insertion): Likewise.
25394         (do_pre_partial_partial_insertion): Likewise.
25396 2018-03-02  Richard Biener  <rguenther@suse.de>
25398         PR target/82005
25399         * config/darwin.c (saved_debug_info_level): New static global.
25400         (darwin_asm_lto_start): Disable debug info generation for LTO out.
25401         (darwin_asm_lto_end): Restore debug info generation settings.
25403 2018-03-01  Martin Liska  <mliska@suse.cz>
25405         PR sanitizer/82484
25406         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
25407         volatile arguments.
25409 2018-03-01  Richard Biener  <rguenther@suse.de>
25411         PR debug/84645
25412         * dwarf2out.c (gen_variable_die): Properly handle late VLA
25413         type annotation with LTO when debug was disabled at compile-time.
25415 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
25417         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
25418         XINT with INTVAL.
25419         (mips_final_postscan_insn): Likewise.
25421 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
25423         PR rtl-optimization/84528
25424         * alias.c (init_alias_target): Add commentary.
25425         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
25426         a unique base value if the frame pointer is not eliminated
25427         to the stack pointer.
25429 2018-03-01  Tom de Vries  <tom@codesourcery.com>
25431         PR rtl-optimization/83327
25432         * lra-int.h (hard_regs_spilled_into): Declare.
25433         * lra.c (hard_regs_spilled_into): Define.
25434         (init_reg_info): Init hard_regs_spilled_into.
25435         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
25436         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
25437         (process_bb_lives): Handle hard_regs_spilled_into.
25438         (lra_create_live_ranges_1): Before doing liveness propagation, clear
25439         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
25441 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
25443         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
25444         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
25445         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
25446         * config/rs6000/aix72.h: New file.
25448 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
25450         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
25451         instead of warning_at with conditional singular and plural messages
25452         where possible.
25454         PR target/52991
25455         * stor-layout.c (update_alignment_for_field): For
25456         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
25457         && !DECL_PACKED (field), do the alignment update, just use
25458         only desired_align instead of MAX (type_align, desired_align)
25459         as the alignment.
25460         (place_field): Don't do known_align < desired_align handling
25461         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
25462         is non-NULL, instead do it after rli->prev_field handling and
25463         only if not within a bitfield word.  For DECL_PACKED (field)
25464         use type_align of BITS_PER_UNIT.
25466 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
25468         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
25469         superfluous parentheses and trailing spaces.
25471 2018-02-28  Richard Biener  <rguenther@suse.de>
25473         PR tree-optimization/84584
25474         * graphite-scop-detection.c (scop_detection::add_scop): Discard
25475         SCoPs with fake exit edge.
25477 2018-02-28  Martin Liska  <mliska@suse.cz>
25479         PR testsuite/84597
25480         * timevar.c (timer::print): Fix format to properly print 100%
25481         values.
25483 2018-02-28  Richard Biener  <rguenther@suse.de>
25485         PR middle-end/84607
25486         * genmatch.c (capture_info::walk_match): Do not mark
25487         captured expressions without operands as expr_p given
25488         they act more like predicates and should be subject to
25489         "lost tail" side-effect preserving.
25491 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
25493         PR rtl-optimization/81611
25494         * auto-inc-dec.c (attempt_change): Move dead note from
25495         mem_insn if it's the next use of regno
25496         (find_address): Take address use of reg holding
25497         non-incremented value.  Add parm to limit search to the named
25498         reg only.
25499         (merge_in_block): Attempt to use a mem insn that is the next
25500         use of the original regno.
25502 2018-02-27  Martin Sebor  <msebor@redhat.com>
25504         PR c++/83871
25505         * doc/invoke.texi (-Wmissing-attributes): New option.
25506         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
25508 2018-02-27  Martin Sebor  <msebor@redhat.com>
25510         PR translation/84207
25511         * diagnostic-core.h (warning_n, error_n, inform_n): Change
25512         n argument to unsigned HOST_WIDE_INT.
25513         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
25514         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
25515         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
25516         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
25518 2018-02-27  Richard Biener  <rguenther@suse.de>
25520         PR tree-optimization/84512
25521         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
25522         Do not use the estimate returned from record_stmt_cost for
25523         the scalar iteration cost but sum properly using add_stmt_cost.
25525 2018-02-27  Richard Biener  <rguenther@suse.de>
25527         PR tree-optimization/84466
25528         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25529         Adjust last change to less strictly validate use operands.
25531 2018-02-27  Martin Liska  <mliska@suse.cz>
25533         PR gcov-profile/84548
25534         * gcov.c (process_file): Allow partial overlap and consider it
25535         also as group functions.
25536         (output_lines): Properly calculate range of lines for a group.
25538 2018-02-27  Martin Liska  <mliska@suse.cz>
25540         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
25541         'ggc' suffixes.  Change first column width.
25542         (timer::print): Fix formatting of the column.
25544 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
25546         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
25547         preserve inline entry blocks for the sake of debug inline
25548         entry point markers alone.
25549         (remove_unused_locals): Suggest in comments a better place to
25550         force the preservation of inline entry blocks that are
25551         otherwise unused, but do not preserve them.
25553 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
25555         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
25557 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
25559         PR target/84039
25560         * config/i386/constraints.md (Bs): Replace
25561         ix86_indirect_branch_register with
25562         TARGET_INDIRECT_BRANCH_REGISTER.
25563         (Bw): Likewise.
25564         * config/i386/i386.md (indirect_jump): Likewise.
25565         (tablejump): Likewise.
25566         (*sibcall_memory): Likewise.
25567         (*sibcall_value_memory): Likewise.
25568         Peepholes of indirect call and jump via memory: Likewise.
25569         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
25570         (*sibcall_value_GOT_32): Likewise.
25571         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25572         (GOT_memory_operand): Likewise.
25573         (call_insn_operand): Likewise.
25574         (sibcall_insn_operand): Likewise.
25575         (GOT32_symbol_operand): Likewise.
25576         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
25578 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
25580         PR rtl-optimization/83496
25581         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
25582         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
25583         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
25584         redundant insn, if any.
25585         (relax_delay_slots): Likewise.
25586         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
25588 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
25590         PR tree-optimization/83965
25591         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
25592         that grouped statements are part of a reduction chain.  Return
25593         true if the statement is not marked as a reduction itself but
25594         is part of a group.
25595         (vect_recog_dot_prod_pattern): Don't check whether the statement
25596         is part of a group here.
25597         (vect_recog_sad_pattern): Likewise.
25598         (vect_recog_widen_sum_pattern): Likewise.
25600 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
25602         PR debug/84545
25603         * final.c (rest_of_clean_state): Also look for calls inside sequences.
25605 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
25607         PR target/84530
25608         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
25609         the bool argument.
25610         (ix86_output_indirect_function_return): New prototype.
25611         (ix86_split_simple_return_pop_internal): Likewise.
25612         * config/i386/i386.c (indirect_return_via_cx): New.
25613         (indirect_return_via_cx_bnd): Likewise.
25614         (indirect_thunk_name): Handle return va CX_REG.
25615         (output_indirect_thunk_function): Create alias for
25616         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
25617         (ix86_output_indirect_jmp): Remove the bool argument.
25618         (ix86_output_indirect_function_return): New function.
25619         (ix86_split_simple_return_pop_internal): Likewise.
25620         * config/i386/i386.md (*indirect_jump): Don't pass false
25621         to ix86_output_indirect_jmp.
25622         (*tablejump_1): Likewise.
25623         (simple_return_pop_internal): Change it to define_insn_and_split.
25624         Call ix86_split_simple_return_pop_internal to split it for
25625         -mfunction-return=.
25626         (simple_return_indirect_internal): Call
25627         ix86_output_indirect_function_return instead of
25628         ix86_output_indirect_jmp.
25630 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
25632         PR bootstrap/84405
25633         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
25634         memset and value initialization afterwards.
25636 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
25638         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
25640 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25642         PR target/84521
25643         * common/config/aarch64/aarch64-common.c
25644         (aarch_option_optimization_table[]): Switch
25645         off fomit-frame-pointer
25647 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25648             Chung-Ju Wu  <jasonwucj@gmail.com>
25650         * config/nds32/nds32-multiple.md (load_multiple): Disallow
25651         volatile memory.
25652         (store_multiple): Ditto.
25654 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25656         * config.gcc: Add --with-cpu support for nds32 target.
25657         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
25658         * config/nds32/nds32.opt: Add -mcpu= option.
25660 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
25662         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
25663         isel=yes): Warn for these deprecated options.
25665 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
25667         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
25668         ISA_2_5_MASKS_EMBEDDED.
25670 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
25672         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
25673         p->max as pointers rather than using iterative_hash_expr.
25675 2018-02-23  Carl Love  <cel@us.ibm.com>
25677         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
25678         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
25679         BU_P8V_OVERLOAD_2.
25680         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
25681         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
25682         P8V_BUILTIN_VEC_VUNSIGNED2.
25684 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
25686         PR target/81572
25687         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25688         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25689         LRA_UNKNOWN_ALT.
25690         * lra-constraints.c (curr_insn_transform): Set up
25691         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
25692         LRA_UNKNOWN_ALT.
25693         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25694         * lra-eliminations.c (spill_pseudos): Ditto.
25695         (process_insn_for_elimination): Ditto.
25696         * lra-lives.c (reg_early_clobber_p): Use the new macros.
25697         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25698         LRA_NON_CLOBBERED_ALT.
25700 2018-02-22  Martin Sebor  <msebor@redhat.com>
25702         PR tree-optimization/84480
25703         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25704         to maybe_diag_stxncpy_trunc.  Call it.
25705         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25706         from gimple_fold_builtin_strcpy.  Print inlining stack.
25707         (handle_builtin_stxncpy): Print inlining stack.
25708         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25710 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
25712         PR target/84176
25713         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25714         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25715         and -fcheck-pointer-bounds are used together.
25716         (indirect_thunk_prefix): New enum.
25717         (indirect_thunk_need_prefix): New function.
25718         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
25719         "_nt" instead of "_bnd" for NOTRACK prefix.
25720         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25721         (output_indirect_thunk_function): Likewise.
25722         (): Likewise.
25723         (ix86_code_end): Update output_indirect_thunk_function calls.
25724         (ix86_output_indirect_branch_via_reg): Replace
25725         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25726         (ix86_output_indirect_branch_via_push): Likewise.
25727         (ix86_output_function_return): Likewise.
25728         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25729         incompatible with -fcf-protection=branch and
25730         -fcheck-pointer-bounds.
25732 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25734         PR target/83335
25735         * config/aarch64/aarch64.c (aarch64_print_address_internal):
25736         Change gcc_assert call to output_operand_lossage.
25738 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25740         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25742 2018-02-22  DJ Delorie  <dj@redhat.com>
25743             Sebastian Perta  <sebastian.perta@renesas.com>
25744             Oleg Endo  <olegendo@gcc.gnu.org>
25746         * config/rx/rx.c (rx_rtx_costs): New function.
25747         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25749 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25751         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25753 2018-02-22  Martin Liska  <mliska@suse.cz>
25755         PR driver/83193
25756         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25757         Add "native" as a possible value.
25759 2018-02-22  Martin Liska  <mliska@suse.cz>
25761         PR driver/83193
25762         * config/i386/i386.c (ix86_option_override_internal):
25763         Add "native" as a possible value for -march and -mtune.
25765 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
25767         PR target/84502
25768         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25769         to all type variants.
25771         PR tree-optimization/84503
25772         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25773         width as info->bitpos + info->bitsize - start.
25774         (merged_store_group::merge_overlapping): Simplify width computation.
25775         (check_no_overlap): New function.
25776         (imm_store_chain_info::try_coalesce_bswap): Compute expected
25777         start + width and last_order of the group, fail if check_no_overlap
25778         fails.
25779         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25780         to group if check_no_overlap fails.
25782 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25784         * config/rs6000/altivec.md: Delete contraint arguments to
25785         define_expand, define_split, and define_peephole2, and in
25786         define_insn_and_split if always unused.
25787         * config/rs6000/darwin.md: Ditto.
25788         * config/rs6000/dfp.md: Ditto.
25789         * config/rs6000/rs6000.md: Ditto.
25790         * config/rs6000/sync.md: Ditto.
25791         * config/rs6000/vector.md: Ditto.
25792         * config/rs6000/vsx.md: Ditto.
25794 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25796         * config/rs6000/altivec.md: Write output control strings as braced
25797         blocks instead of double-quoted strings.
25798         * config/rs6000/darwin.md: Ditto.
25799         * config/rs6000/rs6000.md: Ditto.
25800         * config/rs6000/vector.md: Ditto.
25801         * config/rs6000/vsx.md: Ditto.
25803 2018-02-21  Jason Merrill  <jason@redhat.com>
25805         PR c++/84314 - ICE with templates and fastcall attribute.
25806         * attribs.c (build_type_attribute_qual_variant): Remove assert.
25808 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25810         * ipa-cp.c (determine_versionability): Fix comment typos.
25812 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25814         PR c/84229
25815         * ipa-cp.c (determine_versionability): Do not version functions caling
25816         va_arg_pack.
25818 2018-02-21  Martin Liska  <mliska@suse.cz>
25820         PR driver/83193
25821         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25822         Add "native" as a possible value.
25823         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25824         the macro when native cpu detection is available.
25826 2018-02-21  Martin Liska  <mliska@suse.cz>
25828         PR driver/83193
25829         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25830         Add "native" as a possible value.
25831         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25832         when native cpu detection is available.
25834 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25835             Martin Sebor  <msebor@redhat.com>
25837         PR tree-optimization/84478
25838         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25839         false.
25840         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25841         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25842         support which is conservatively correct, for 2 only stay conservative
25843         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25844         argument to the 2 argument get_range_strlen, adjust 6 arg
25845         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25846         false.
25847         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25848         (gimple_fold_builtin_strlen): Pass true as last argument to
25849         get_range_strlen.
25851 2018-02-20  Martin Sebor  <msebor@redhat.com>
25853         PR middle-end/84095
25854         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25855         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25856         (builtin_memref::builtin_memref): Factor out parts into
25857         set_base_and_offset and call it.
25859 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25861         PR middle-end/84406
25862         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25863         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25864         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25865         search at the associated MODE_INT.
25867 2018-02-20  Jeff Law  <law@redhat.com>
25869         PR middle-end/82123
25870         PR tree-optimization/81592
25871         PR middle-end/79257
25872         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25873         for range data rather than using global data.
25874         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25875         range data rather than using global data.
25876         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25877         pass it to children as needed.
25878         (struct directive::fmtresult): Similarly.
25879         (struct directive::set_width): Similarly.
25880         (struct directive::set_precision): Similarly.
25881         (format_integer, format_directive, parse_directive): Similarly.
25882         (format_none): Accept unnamed vr_values parameter.
25883         (format_percent, format_floating, format_character): Similarly.
25884         (format_string, format_plain): Similarly.
25885         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25886         the EVRP range analyzer for range data rather than using global data.
25887         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25888         gimple-ssa-evrp-analyze.h
25889         (class sprintf_dom_walker): Add after_dom_children member function.
25890         Add evrp_range_analyzer member.
25891         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25892         range analyzer as needed.
25893         (sprintf_dom_walker::after_dom_children): New member function.
25894         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25895         if not optimizing.
25896         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25897         (evrp_range_analyzer::pop_to_marker): Likewise.
25899 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25901         PR tree-optimization/84419
25902         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25903         with the required type if its current type is compatible but
25904         different.
25906 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25908         PR middle-end/82004
25909         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25910         after vectorization.
25912 2018-02-20  Martin Liska  <mliska@suse.cz>
25914         PR driver/83193
25915         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25916         possible values if we don't have a hint.
25918 2018-02-20  Martin Liska  <mliska@suse.cz>
25920         PR c/84310
25921         PR target/79747
25922         * final.c (shorten_branches): Build align_tab array with one
25923         more element.
25924         * opts.c (finish_options): Add alignment option limit check.
25925         (MAX_CODE_ALIGN): Likewise.
25926         (MAX_CODE_ALIGN_VALUE): Likewise.
25927         * doc/invoke.texi: Document maximum allowed option value for
25928         all -falign-* options.
25930 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25932         PR target/84146
25933         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25934         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25935         * var-tracking.c (emit_note_insn_var_location): Remove all references
25936         to NOTE_INSN_CALL_ARG_LOCATION.
25937         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25938         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25939         Use copy_rtx_if_shared.
25940         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25941         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25942         (dwarf2out_var_location): Remove handling of
25943         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25944         on call_insn.
25945         * final.c (final_scan_insn): Remove all references to
25946         NOTE_INSN_CALL_ARG_LOCATION.
25947         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25948         before dumping final insns.
25949         * except.c (emit_note_eh_region_end): Remove all references to
25950         NOTE_INSN_CALL_ARG_LOCATION.
25951         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25952         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25953         * config/arc/arc.c (hwloop_optimize): Likewise.
25954         * config/arm/arm.c (create_fix_barrier): Likewise.
25955         * config/s390/s390.c (s390_chunkify_start): Likewise.
25956         * config/sh/sh.c (find_barrier): Likewise.
25957         * config/i386/i386.c (rest_of_insert_endbranch,
25958         ix86_seh_fixup_eh_fallthru): Likewise.
25959         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25960         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25961         * config/frv/frv.c (frv_function_prologue): Likewise.
25962         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25963         reg note.
25964         (note_outside_basic_block_p): Remove all references to
25965         NOTE_INSN_CALL_ARG_LOCATION.
25966         * gengtype.c (adjust_field_rtx_def): Likewise.
25967         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25968         Likewise.
25969         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25970         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25972         PR c++/84444
25973         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25974         is ADDR_EXPR.
25976         PR tree-optimization/84452
25977         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25978         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25979         is NULL.
25981 2018-02-19  Martin Liska  <mliska@suse.cz>
25983         PR sanitizer/82183
25984         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25986 2018-02-19  Martin Liska  <mliska@suse.cz>
25987             Richard Sandiford  <richard.sandiford@linaro.org>
25989         PR tree-optimization/82491
25990         * gimple-fold.c (get_base_constructor): Make earlier bail out
25991         to prevent ubsan.
25993 2018-02-19  Carl Love  <cel@us.ibm.com>
25995         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25996         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25997         BU_P8V_OVERLOAD_1.
25998         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25999         P8V_BUILTIN_VEC_NEG.
26001 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
26003         * config/rl78/rl78.md (movdf): New define expand.
26005 2018-02-19  Martin Liska  <mliska@suse.cz>
26007         PR other/80589
26008         * doc/invoke.texi: Fix typo.
26009         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
26011 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
26013         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
26014         handle rs6000_single_float and rs6000_double_float specially for
26015         e500 family CPUs.
26017 2018-02-16  Jeff Law  <law@redhat.com>
26019         * config/rx/rx.c (add_pop_cfi_notes): New function.;
26020         (pop_regs): Use it.
26022 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
26024         PR ipa/84425
26025         * ipa-inline.c (inline_small_functions): Fix a typo.
26027 2018-02-16  Nathan Sidwell  <nathan@acm.org>
26029         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
26031 2018-02-16  Carl Love  <cel@us.ibm.com>
26033         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
26034         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
26035         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
26036         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
26037         expansion to P8V_BUILTIN_VEC_FLOAT2.
26039 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
26041         PR rtl-optimization/70023
26042         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
26043         src_regno into account.
26045 2018-02-16  Carl Love  <cel@us.ibm.com>
26047         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
26048         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
26049         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
26050         * config/rs6000/rs6000.c: Remove case statements for
26051         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
26052         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
26053         and P9V_BUILTIN_VEC_VINSERT4B.
26054         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
26055         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
26056         * config/rs6000/vsx.md:
26057         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
26058         vec_insert4b.
26060 2018-02-16  Carl Love  <cel@us.ibm.com>
26062         * config/rs6000/altivec.h: Add builtin names vec_extract4b
26063         vec_insert4b.
26064         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
26065         definitions.
26066         * config/rs6000/rs6000-c.c: Add the definitions for
26067         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
26068         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
26069         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
26070         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
26071         definition for insert4b and define insn *insert3b_internal.
26072         * doc/extend.texi: Add documentation for vec_extract4b.
26074 2018-02-16  Nathan Sidwell  <nathan@acm.org>
26076         * doc/extend.texi (Backwards Compatibility): Mention friend
26077         injection.  Note for-scope is deprecated.
26078         * doc/invoke.texi (-ffriend-injection): Deprecate.
26080 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
26082         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
26083         that moved to I2, also allow destinations that are a paradoxical
26084         subreg (instead of a normal reg).
26086 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
26088         PR target/83831
26089         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
26090         to QImode.
26092 2018-02-16  Richard Biener  <rguenther@suse.de>
26094         PR tree-optimization/84037
26095         PR tree-optimization/84016
26096         PR target/82862
26097         * config/i386/i386.c (ix86_builtin_vectorization_cost):
26098         Adjust vec_construct for the fact we need additional higher latency
26099         128bit inserts for AVX256 and AVX512 vector builds.
26100         (ix86_add_stmt_cost): Scale vector construction cost for
26101         elementwise loads.
26103 2018-02-16  Richard Biener  <rguenther@suse.de>
26105         PR tree-optimization/84417
26106         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
26107         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
26108         (non_rewritable_lvalue_p): Likewise, use poly-ints.
26110 2018-02-16  Martin Liska  <mliska@suse.cz>
26112         PR sanitizer/84307
26113         * internal-fn.def (ASAN_CHECK): Set proper flags.
26114         (ASAN_MARK): Likewise.
26116 2018-02-16  Julia Koval  <julia.koval@intel.com>
26118         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
26119         from PTA_CANNONLAKE.
26121 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
26123         PR target/84272
26124         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
26125         Use ++iter rather than iter++ for std::list iterators.
26126         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
26127         defer deleting them until all nodes in the forest are processed.  Do
26128         free even leaf nodes.  Change to_process into auto_vec.
26130         PR bootstrap/84405
26131         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
26132         * vec.h (vec_default_construct): Use memset instead of placement new
26133         if BROKEN_VALUE_INITIALIZATION is defined.
26134         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
26135         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
26136         is defined.
26138         PR rtl-optimization/83723
26139         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
26140         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
26141         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
26142         recursive calls.
26143         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
26144         callers.
26145         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
26147 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
26149         PR rtl-optimization/81443
26150         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
26151         from inner REGs to paradoxical SUBREGs.
26153 2018-02-16  Richard Biener  <rguenther@suse.de>
26155         PR tree-optimization/84399
26156         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
26157         For operands we can analyze at their definition make sure we can
26158         analyze them at each use as well.
26160 2018-02-16  Richard Biener  <rguenther@suse.de>
26162         PR tree-optimization/84190
26163         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
26164         volatile accesses if the decl isn't volatile.
26166 2018-02-15  Jason Merrill  <jason@redhat.com>
26168         PR c++/84314 - ICE with templates and fastcall attribute.
26169         * attribs.c (build_type_attribute_qual_variant): Don't clobber
26170         TYPE_CANONICAL on an existing type.
26172 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
26174         PR tree-optimization/84383
26175         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
26176         dstoff nor call operand_equal_p if dstbase is NULL.
26178         PR tree-optimization/84334
26179         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
26180         also a CONSTANT_CLASS_P, punt.
26182 2018-02-14  Jim Wilson  <jimw@sifive.com>
26184         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
26185         first SMALL_OPERAND check.  New local min_second_step.  Move assert
26186         to where locals are set.  Add TARGET_RVC support.
26187         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
26189 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
26191         * doc/invoke.texi: Correct -Wformat-overflow code sample.
26193 2018-02-14  Martin Sebor  <msebor@redhat.com>
26195         PR tree-optimization/83698
26196         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
26197         arrays constrain the offset range to their bounds.
26198         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
26199         (builtin_access::overlap): Avoid setting the size of overlap if it's
26200         already been set.
26201         (maybe_diag_overlap): Also consider arrays when deciding what values
26202         of offsets to include in diagnostics.
26204 2018-02-14  Martin Sebor  <msebor@redhat.com>
26206         PR c/84108
26207         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
26208         that correspond to the kind of a declaration.
26210 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
26212         PR target/83984
26213         * config/pa/pa.md: Load address of PIC label using the linkage table
26214         if the label is nonlocal.
26216 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26218         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
26219         warning message if user requests -maltivec=be.
26220         * doc/invoke.texi: Document deprecation of -maltivec=be.
26222 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
26224         PR target/84220
26225         * config/rs6000/rs6000-c.c: Update definitions for
26226         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
26227         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
26229 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26231         PR target/84239
26232         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
26233         add _get_ssp intrinsics. Remove argument from
26234         __builtin_ia32_rdssp[d|q].
26235         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
26236         * config/i386/i386-builtin.def: Remove argument from
26237         __builtin_ia32_rdssp[d|q].
26238         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
26239         ix86_expand_special_args_builtin for _rdssp[d|q].
26240         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
26241         Clear register before usage.
26242         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
26243         Add documentation for new _get_ssp and _inc_ssp intrinsics.
26245 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
26247         PR tree-optimization/84357
26248         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
26249         operand 1 of an ARRAY_REF too.
26251 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
26253         PR target/83831
26254         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
26255         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
26256         declarations.
26257         (set_of_reg): New struct.
26258         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
26259         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
26260         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
26261         functions.
26262         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
26263         Split into bitclr, bitset, bitinvert patterns if appropriate.
26264         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
26265         use rx_fuse_in_memory_bitop.
26266         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
26267         to named insn, correct maximum insn length.
26269 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
26271         PR target/79242
26272         * machmode.def: Define a complex mode for PARTIAL_INT.
26273         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
26274         MODE_PARTIAL_INT.
26275         * doc/rtl.texi: Document CSPImode.
26276         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
26277         handling.
26278         (msp430_hard_regno_nregs_with_padding): Likewise.
26280 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
26282         PR target/84279
26283         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
26285 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
26287         PR rtl-optimization/84169
26288         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
26289         we generated a parallel as new i3 and we split that to new i2 and i3
26290         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
26291         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
26292         those to i2, not i1.  Partially rewrite this scan code.
26294 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
26296         PR c/82210
26297         * stor-layout.c (place_field): For variable length fields, adjust
26298         offset_align afterwards not just based on the field's alignment,
26299         but also on the size.
26301         PR middle-end/84309
26302         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
26303         of exps and logs in the use_exp2 case.
26305 2018-02-13  Jeff Law  <law@redhat.com>
26307         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
26308         entry for "vector".
26310         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
26311         ARGS as unused.
26313 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
26315         PR debug/84342
26316         PR debug/84319
26317         * common.opt (gas-loc-support, gas-locview-support): New.
26318         (ginline-points, ginternal-reset-location-views): New.
26319         * doc/invoke.texi: Document them.  Use @itemx where intended.
26320         (gvariable-location-views): Adjust.
26321         * target.def (reset_location_view): New.
26322         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
26323         (TARGET_RESET_LOCATION_VIEW): New.
26324         * doc/tm.texi: Rebuilt.
26325         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
26326         (dwarf2out_default_as_locview_support): New.
26327         (output_asm_line_debug_info): Use option variables.
26328         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
26329         (output_loc_list): Likewise.
26330         (add_high_low_attributes): Check option variables.
26331         Don't output entry view attribute in strict mode.
26332         (gen_inlined_subroutine_die): Check option variables.
26333         (dwarf2out_inline_entry): Likewise.
26334         (init_sections_and_labels): Likewise.
26335         (dwarf2out_early_finish): Likewise.
26336         (maybe_reset_location_view): New, from...
26337         (dwarf2out_var_location): ... here.  Call it.
26338         * debug.h (dwarf2out_default_as_loc_support): Declare.
26339         (dwarf2out_default_as_locview_support): Declare.
26340         * hooks.c (hook_int_rtx_insn_0): New.
26341         * hooks.h (hook_int_rtx_insn_0): Declare.
26342         * toplev.c (process_options): Take -gas-loc-support and
26343         -gas-locview-support from dwarf2out.  Enable
26344         -gvariable-location-views by default only with locview
26345         assembler support.  Enable -ginternal-reset-location-views by
26346         default only if the target defines the corresponding hook.
26347         Enable -ginline-points by default if location views are
26348         enabled; force it disabled if statement frontiers are
26349         disabled.
26350         * tree-inline.c (expand_call_inline): Check option variables.
26351         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
26353 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
26355         PR tree-optimization/84321
26356         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
26357         handling.  Also check whether the anti-range contains any values
26358         that satisfy the mask; switch to a VR_RANGE if not.
26360 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
26362         PR sanitizer/84340
26363         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
26365 2018-02-13  Martin Jambor  <mjambor@suse.cz>
26367         PR c++/83990
26368         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
26369         of call statements, also set location of a load to a temporary.
26371 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
26373         * config/rl78/rl78.c (add_vector_labels): New function.
26374         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
26375         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
26376         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
26377         which checks that no arguments are passed.
26378         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
26379         * doc/extend.texi: Documentation for the new attribute.
26381 2018-02-13  Andreas Schwab  <schwab@suse.de>
26383         * config/riscv/linux.h (CPP_SPEC): Define.
26385 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
26387         PR target/84335
26388         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
26389         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
26390         OPTION_MASK_ISA_AES as first argument to def_builtin_const
26391         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
26392         instead of OPTION_MASK_ISA_PCLMUL as first argument to
26393         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
26394         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
26395         temporarily for AES and PCLMUL builtins.
26397         PR tree-optimization/84339
26398         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
26399         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
26400         Formatting fixes.
26402         PR middle-end/84309
26403         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
26404         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
26405         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
26406         inline function.
26407         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
26408         inline function.
26409         * omp-simd-clone.h: New file.
26410         * omp-simd-clone.c: Include omp-simd-clone.h.
26411         (expand_simd_clones): No longer static.
26412         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
26413         cgraph.h and omp-simd-clone.h.
26414         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
26415         (vect_recog_widen_shift_pattern): Formatting fix.
26416         (vect_pattern_recog_1): Don't check optab for calls.
26418         PR target/84336
26419         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
26420         operands[2] into a REG before using gen_lowpart on it.
26422 2018-02-12  Jeff Law  <law@redhat.com>
26424         PR target/83760
26425         * config/sh/sh.c (find_barrier): Consider a sibling call
26426         a barrier as well.
26428         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
26429         successfully back substituting a reg.
26431 2018-02-12  Richard Biener  <rguenther@suse.de>
26433         PR tree-optimization/84037
26434         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
26435         parameter, move visited init to caller.
26436         (vect_slp_analyze_operations): Separate cost from validity
26437         check, initialize visited once for all instances.
26438         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
26439         for all instances.
26440         * tree-vect-stmts.c (vect_model_simple_cost): Make early
26441         out an assert.
26442         (vect_model_promotion_demotion_cost): Likewise.
26443         (vectorizable_bswap): Guard cost modeling with !slp_node
26444         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
26445         SLP stmts.
26446         (vectorizable_call): Likewise.
26447         (vectorizable_conversion): Likewise.
26448         (vectorizable_assignment): Likewise.
26449         (vectorizable_shift): Likewise.
26450         (vectorizable_operation): Likewise.
26451         (vectorizable_store): Likewise.
26452         (vectorizable_load): Likewise.
26453         (vectorizable_condition): Likewise.
26454         (vectorizable_comparison): Likewise.
26456 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
26458         PR sanitizer/84307
26459         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
26460         (ASAN_MARK): Fix fnspec to account for return value, change pointer
26461         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
26463 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26465         PR middle-end/83665
26466         * params.def (inline-min-speedup): Increase from 8 to 15.
26467         (max-inline-insns-auto): Decrease from 40 to 30.
26468         * ipa-split.c (consider_split): Add some buffer for function to
26469         be considered inlining candidate.
26470         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
26471         default values.
26473 2018-02-12  Richard Biener  <rguenther@suse.de>
26475         PR tree-optimization/84037
26476         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
26477         matched stmts if we cannot swap the non-matched ones.
26479 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
26481         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
26482         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
26483         _mm_maskz_scalef_round_ss): New intrinsics.
26484         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
26485         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
26486         __builtin_ia32_scalefss_round): Remove.
26487         (__builtin_ia32_scalefsd_mask_round,
26488         __builtin_ia32_scalefss_mask_round): New intrinsics.
26489         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
26490         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
26491         ((match_operand:VF_128 2 "<round_nimm_predicate>"
26492         "<round_constraint>")): Changed to ...
26493         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
26494         "<round_scalar_constraint>")): ... this.
26495         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
26496         %0, %1, %2<round_op3>}"): Changed to ...
26497         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
26498         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
26499         %2<round_scalar_mask_op3>}"): ... this.
26500         * config/i386/subst.md (round_scalar_nimm_predicate): New.
26502 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
26504         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
26505         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
26506         (_mm_maskz_sqrt_round_ss): New intrinsics.
26507         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
26508         (__builtin_ia32_sqrtsd_mask_round)
26509         (__builtin_ia32_sqrtss_mask_round): New builtins.
26510         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
26511         (__builtin_ia32_sqrtss_round): Remove.
26512         (__builtin_ia32_sqrtsd_mask_round)
26513         (__builtin_ia32_sqrtss_mask_round): New builtins.
26514         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
26515         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
26516         ((match_operand:VF_128 1 "vector_operand"
26517         "xBm,<round_constraint>")): Changed to ...
26518         ((match_operand:VF_128 1 "vector_operand"
26519         "xBm,<round_scalar_constraint>")): ... this.
26520         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
26521         %0, %2, %<iptr>1<round_op3>}): Changed to ...
26522         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
26523         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
26524         %<iptr>1<round_scalar_mask_op3>}): ... this.
26525         ((set_attr "prefix" "<round_prefix>")): Changed to ...
26526         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
26528 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
26530         PR target/84266
26531         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
26532         Cast vec_cmpeq result to correct type.
26533         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
26534         Cast vec_cmpgt result to correct type.
26536 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
26538         * final.c (final_scan_insn_1): Renamed from...
26539         (final_scan_insn): ... this.  New wrapper, to recover
26540         seen from the outermost call in recursive ones.
26541         * config/sparc/sparc.c (output_return): Drop seen from call.
26542         (output_sibcall): Likewise.
26543         * config/visium/visium.c (output_branch): Likewise.
26545 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
26547         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
26548         function label.
26550 2018-02-10  Alan Modra  <amodra@gmail.com>
26552         PR target/84300
26553         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
26554         Specify LR as an input.
26556 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
26558         PR sanitizer/83987
26559         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
26560         remove_member_access_dummy_vars): New functions.
26561         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
26562         lower_omp_1, execute_lower_omp): Use them.
26564         PR rtl-optimization/84308
26565         * shrink-wrap.c (spread_components): Release todo vector.
26567 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
26569         PR rtl-optimization/57193
26570         * ira-color.c (struct allocno_color_data): Add member
26571         conflict_allocno_hard_prefs.
26572         (update_conflict_allocno_hard_prefs): New.
26573         (bucket_allocno_compare_func): Add a preference based on
26574         conflict_allocno_hard_prefs.
26575         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
26576         (color_allocnos): Remove a dead code.  Initiate
26577         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
26579 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
26581         PR target/84226
26582         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
26583         constraint from =wa to wa.  Avoid a subreg on the output operand,
26584         instead use a pseudo and subreg it in a move.
26585         (p9_xxbrd_<mode>): Changed to ...
26586         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
26587         (p9_xxbrd_v2df): New expander.
26588         (p9_xxbrw_<mode>): Changed to ...
26589         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
26590         (p9_xxbrw_v4sf): New expander.
26592 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
26594         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
26596 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
26598         PR target/83926
26599         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
26600         multiply in 32-bit mode.
26601         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
26602         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
26603         mode.
26605 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
26607         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
26608         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
26609         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
26610         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
26612 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
26614         PR lto/84213
26615         * dwarf2out.c (is_trivial_indirect_ref): New function.
26616         (dwarf2out_late_global_decl): Do not generate a location
26617         attribute for variables that have a non-trivial DECL_VALUE_EXPR
26618         and that are not defined in the current unit.
26620 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
26622         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
26623         instead of a libcall for UNORDERED.
26625 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
26627         PR target/82641
26628         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
26629         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
26631 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26633         PR target/PR84295
26634         * config/s390/s390.c (s390_set_current_function): Invoke
26635         s390_indirect_branch_settings also if fndecl didn't change.
26637 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26639         * config/rs6000/rs6000.md (blockage): Set length to zero.
26641 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
26643         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
26645 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
26647         PR sanitizer/84285
26648         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
26649         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
26650         -static-lib*san.
26652         PR debug/84252
26653         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
26654         PARALLEL incoming that failed vt_get_decl_and_offset check.
26656         PR middle-end/84237
26657         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
26658         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
26659         TREE_READONLY bit.
26660         (get_variable_section): For decls in named .bss* sections pass true as
26661         second argument to bss_initializer_p.
26663 2018-02-09  Marek Polacek  <polacek@redhat.com>
26664             Jakub Jelinek  <jakub@redhat.com>
26666         PR c++/83659
26667         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
26668         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
26669         Sync some changes from cxx_fold_indirect_ref.
26671 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26673         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
26674         markers.
26675         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
26676         (BLOCK_INLINE_ENTRY_LABEL): New.
26677         (dwarf2out_var_location): Disregard inline entry markers.
26678         (inline_entry_data): New struct.
26679         (inline_entry_data_hasher): New hashtable type.
26680         (inline_entry_data_hasher::hash): New.
26681         (inline_entry_data_hasher::equal): New.
26682         (inline_entry_data_table): New variable.
26683         (add_high_low_attributes): Add DW_AT_entry_pc and
26684         DW_AT_GNU_entry_view attributes if a pending entry is found
26685         in inline_entry_data_table.  Add old entry_pc attribute only
26686         if debug nonbinding markers are disabled.
26687         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26688         markers are enabled.
26689         (block_within_block_p, dwarf2out_inline_entry): New.
26690         (dwarf2out_finish): Check that no entries remained in
26691         inline_entry_data_table.
26692         * final.c (reemit_insn_block_notes): Handle inline entry notes.
26693         (final_scan_insn, notice_source_line): Likewise.
26694         (rest_of_clean_state): Skip inline entry markers.
26695         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26696         markers.
26697         * gimple.c (gimple_build_debug_inline_entry): New.
26698         * gimple.h (enum gimple_debug_subcode): Add
26699         GIMPLE_DEBUG_INLINE_ENTRY.
26700         (gimple_build_debug_inline_entry): Declare.
26701         (gimple_debug_inline_entry_p): New.
26702         (gimple_debug_nonbind_marker_p): Adjust.
26703         * insn-notes.def (INLINE_ENTRY): New.
26704         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26705         inline entry marker notes.
26706         (print_insn): Likewise.
26707         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26708         (INSN_DEBUG_MARKER_KIND): Likewise.
26709         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26710         * tree-inline.c (expand_call_inline): Build and insert
26711         debug_inline_entry stmt.
26712         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26713         inline entry blocks early, if nonbind markers are enabled.
26714         (dump_scope_block): Dump fragment info.
26715         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26716         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26717         (gimple_build_debug_inline_entry): New.
26718         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26719         Enable/disable inline entry points too.
26720         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26721         (DEBUG_INSN): Describe inline entry markers.
26723         * common.opt (gvariable-location-views): New.
26724         (gvariable-location-views=incompat5): New.
26725         * config.in: Rebuilt.
26726         * configure: Rebuilt.
26727         * configure.ac: Test assembler for view support.
26728         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26729         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26730         * dwarf2out.c (var_loc_view): New typedef.
26731         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26732         (dwarf2out_locviews_in_attribute): New.
26733         (dwarf2out_locviews_in_loclist): New.
26734         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26735         (enum dw_line_info_opcode): Add LI_adv_address.
26736         (struct dw_line_info_table): Add view.
26737         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26738         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26739         (zero_view_p): New variable.
26740         (ZERO_VIEW_P): New macro.
26741         (output_asm_line_debug_info): New.
26742         (struct var_loc_node): Add view.
26743         (add_AT_view_list, AT_loc_list): New.
26744         (add_var_loc_to_decl): Add view param.  Test it against last.
26745         (new_loc_list): Add view params.  Record them.
26746         (AT_loc_list_ptr): Handle loc and view lists.
26747         (view_list_to_loc_list_val_node): New.
26748         (print_dw_val): Handle dw_val_class_view_list.
26749         (size_of_die): Likewise.
26750         (value_format): Likewise.
26751         (loc_list_has_views): New.
26752         (gen_llsym): Set vl_symbol too.
26753         (maybe_gen_llsym, skip_loc_list_entry): New.
26754         (dwarf2out_maybe_output_loclist_view_pair): New.
26755         (output_loc_list): Output view list or entries too.
26756         (output_view_list_offset): New.
26757         (output_die): Handle dw_val_class_view_list.
26758         (output_dwarf_version): New.
26759         (output_compilation_unit_header): Use it.
26760         (output_skeleton_debug_sections): Likewise.
26761         (output_rnglists, output_line_info): Likewise.
26762         (output_pubnames, output_aranges): Update version comments.
26763         (output_one_line_info_table): Output view numbers in asm comments.
26764         (dw_loc_list): Determine current endview, pass it to new_loc_list.
26765         Call maybe_gen_llsym.
26766         (loc_list_from_tree_1): Adjust.
26767         (add_AT_location_description): Create view list attribute if
26768         needed, check it's absent otherwise.
26769         (convert_cfa_to_fb_loc_list): Adjust.
26770         (maybe_emit_file): Call output_asm_line_debug_info for test.
26771         (dwarf2out_var_location): Reset views as needed.  Precompute
26772         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
26773         attribute.  Set view.
26774         (new_line_info_table): Reset next view.
26775         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26776         (dwarf2out_source_line): Likewise.  Output view resets and labels to
26777         the assembler, or select appropriate line info opcodes.
26778         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26779         (optimize_string_length): Catch it.  Adjust.
26780         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
26781         dw_val_class_view_list, and remove it if no longer needed.
26782         (hash_loc_list): Hash view numbers.
26783         (loc_list_hasher::equal): Compare them.
26784         (optimize_location_lists): Check whether a view list symbol is
26785         needed, and whether the locview attribute is present, and
26786         whether they match.  Remove the locview attribute if no longer
26787         needed.
26788         (index_location_lists): Call skip_loc_list_entry for test.
26789         (dwarf2out_finish): Call output_asm_line_debug_info for test.
26790         Use output_dwarf_version.
26791         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26792         (struct dw_val_node): Add val_view_list.
26793         * final.c (SEEN_NEXT_VIEW): New.
26794         (set_next_view_needed): New.
26795         (clear_next_view_needed): New.
26796         (maybe_output_next_view): New.
26797         (final_start_function): Rename to...
26798         (final_start_function_1): ... this.  Take pointer to FIRST,
26799         add SEEN parameter.  Emit param bindings in the initial view.
26800         (final_start_function): Reintroduce SEEN-less interface.
26801         (final): Rename to...
26802         (final_1): ... this.  Take SEEN parameter.  Output final pending
26803         next view at the end.
26804         (final): Reintroduce seen-less interface.
26805         (final_scan_insn): Output pending next view before switching
26806         sections or ending a block.  Mark the next view as needed when
26807         outputting variable locations.  Notify debug backend of section
26808         changes, and of location view changes.
26809         (rest_of_handle_final): Adjust.
26810         * toplev.c (process_options): Autodetect value for debug variable
26811         location views option.  Warn on incompat5 without -gdwarf-5.
26812         * doc/invoke.texi (gvariable-location-views): New.
26813         (gvariable-location-views=incompat5): New.
26814         (gno-variable-location-views): New.
26816 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26818         PR tree-optimization/84136
26819         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26820         that the result of find_edge is non-NULL.
26822 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26824         PR target/83008
26825         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26826         storing integer register in SImode.  Fix cost of 256 and 512
26827         byte aligned SSE register store.
26829 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26831         * config/i386/i386.c (ix86_multiplication_cost): Fix
26832         multiplication cost for TARGET_AVX512DQ.
26834 2018-02-08  Marek Polacek  <polacek@redhat.com>
26836         PR tree-optimization/84238
26837         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26838         get_range_strlen.
26840 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26842         PR tree-optimization/84265
26843         * tree-vect-stmts.c (vectorizable_store): Don't treat
26844         VMAT_CONTIGUOUS accesses as grouped.
26845         (vectorizable_load): Likewise.
26847 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26849         PR tree-optimization/81635
26850         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26851         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26852         (test_round_for_mask): New functions.
26853         (wide_int_cc_tests): Call test_round_for_mask.
26854         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26855         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26856         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26857         range returned by get_range_info.
26859 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26861         PR ipa/81360
26862         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26863         * symtab.c: Include builtins.h
26864         (symtab_node::output_to_lto_symbol_table_p): Move here
26865         from lto-streamer-out.c:output_symbol_p.
26866         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26867         (output_symbol_p): Move all logic to symtab.c
26868         (produce_symtab): Update.
26870 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26872         * config/s390/s390-opts.h (enum indirect_branch): Define.
26873         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26874         (s390_indirect_branch_via_thunk)
26875         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26876         (enum s390_indirect_branch_type): Define.
26877         * config/s390/s390.c (struct s390_frame_layout, struct
26878         machine_function): Remove.
26879         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26880         (indirect_branch_table_label_no, indirect_branch_table_name):
26881         Define variables.
26882         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26883         (enum s390_indirect_branch_option): Define.
26884         (s390_return_addr_from_memory): New function.
26885         (s390_handle_string_attribute): New function.
26886         (s390_attribute_table): Add new attribute handler.
26887         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26888         (s390_indirect_branch_via_thunk): New function.
26889         (s390_indirect_branch_via_inline_thunk): New function.
26890         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26891         sibling call optimization for non z10 compiles.
26892         (s390_emit_call): Force indirect branch target to be a single
26893         register.  Add r1 clobber for non-z10 compiles.
26894         (s390_emit_epilogue): Emit return jump via return_use expander.
26895         (s390_reorg): Handle JUMP_INSNs as execute targets.
26896         (s390_option_override_internal): Perform validity checks for the
26897         new command line options.
26898         (s390_indirect_branch_attrvalue): New function.
26899         (s390_indirect_branch_settings): New function.
26900         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26901         (s390_output_indirect_thunk_function):  New function.
26902         (s390_code_end): Implement target hook.
26903         (s390_case_values_threshold): Implement target hook.
26904         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26905         macros.
26906         * config/s390/s390.h (struct s390_frame_layout)
26907         (struct machine_function): Move here from s390.c.
26908         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26909         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26910         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26911         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26912         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26913         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26914         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26915         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26916         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26917         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26918         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26919         (mnemonic attribute): Add values which aren't recognized
26920         automatically.
26921         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26922         pattern for branch conversion.  Fix mnemonic attribute.
26923         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26924         indirect branch via thunk if requested.
26925         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26926         ("*indirect_jump"): Disable for branch conversion using out of
26927         line thunks.
26928         ("indirect_jump_via_thunk<mode>_z10")
26929         ("indirect_jump_via_thunk<mode>")
26930         ("indirect_jump_via_inlinethunk<mode>_z10")
26931         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26932         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26933         ("casesi_jump_via_inlinethunk<mode>_z10")
26934         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26935         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26936         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26937         ("*indirect2_jump"): Disable for branch conversion.
26938         ("casesi_jump"): Turn into expander and expand patterns for branch
26939         conversion.
26940         ("return_use"): New expander.
26941         ("*return"): Emit return via thunk and rename it to ...
26942         ("*return<mode>"): ... this one.
26943         * config/s390/s390.opt: Add new options and and enum for the
26944         option values.
26946 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26948         * lra-constraints.c (match_reload): Unconditionally use
26949         gen_lowpart_SUBREG, rather than selecting between that
26950         and equivalent gen_rtx_SUBREG code.
26952 2018-02-08  Richard Biener  <rguenther@suse.de>
26954         PR tree-optimization/84233
26955         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26956         changed flag instead of boguously re-using phi_inserted.
26958 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26960         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26961         static local variables.
26963 2018-02-08  Richard Biener  <rguenther@suse.de>
26965         PR tree-optimization/84278
26966         * tree-vect-stmts.c (vectorizable_store): When looking for
26967         smaller vector types to perform grouped strided loads/stores
26968         make sure the mode is supported by the target.
26969         (vectorizable_load): Likewise.
26971 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26973         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26974         Increase LDP/STP opportunities by adding adjacent callee-saves.
26976 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26978         PR rtl-optimization/84068
26979         PR rtl-optimization/83459
26980         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26982 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26984         PR tree-optimization/84224
26985         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26986         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26987         non-zero arguments.
26989 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26991         PR target/84113
26992         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26993         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26994         count, remove one USE.
26996 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26998         * doc/install.texi (Configuration): Document the
26999         --with-long-double-format={ibm,ieee} PowerPC configuration
27000         options.
27002         PR target/84154
27003         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
27004         Convert from define_expand to be define_insn_and_split.  Rework
27005         float/double/_Float128 conversions to QI/HI/SImode to work with
27006         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
27007         conversions to QI/HImode types did a store and then a load to
27008         truncate the value.  For conversions to VSX registers, don't split
27009         the insn, instead emit the code directly.  Use the code iterator
27010         any_fix to combine signed and unsigned conversions.
27011         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
27012         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
27013         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27014         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
27015         (fix_<mode>di2_hw): Likewise.
27016         (fixuns_<mode>di2_hw): Likewise.
27017         (fix_<mode>si2_hw): Likewise.
27018         (fixuns_<mode>si2_hw): Likewise.
27019         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
27020         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27021         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
27022         fix<uns>_trunc<SFDF:mode>si2_p8.
27023         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
27024         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
27025         (fix<uns>_<mode>_mem): Likewise.
27026         (fctiw<u>z_<mode>_mem): Likewise.
27027         (fix<uns>_<mode>_mem): Likewise.
27028         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
27029         the register allocator from doing a direct move to the GPRs to do
27030         a store, and instead use the ISA 3.0 store byte/half-word from
27031         vector register instruction.  For IEEE 128-bit floating point,
27032         also optimize stores of 32-bit ints.
27033         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
27035 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
27037         * genextract.c (push_pathstr_operand): New function to support
27038         [a-zA-Z].
27039         (walk_rtx): Call push_pathstr_operand.
27040         (print_path): Support [a-zA-Z].
27042 2018-02-07  Richard Biener  <rguenther@suse.de>
27044         PR tree-optimization/84037
27045         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
27046         (cse_and_gimplify_to_preheader): Declare.
27047         (vect_get_place_in_interleaving_chain): Likewise.
27048         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27049         ivexpr_map.
27050         (_loop_vec_info::~_loop_vec_info): Delete it.
27051         (cse_and_gimplify_to_preheader): New function.
27052         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
27053         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
27054         (vectorizable_load): Likewise.  For grouped stores always base
27055         the IV on the first element.
27056         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
27057         condition before gimplifying.
27059 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
27061         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
27062         *DIV_EXPR and *MOD_EXPR.
27064 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
27066         PR target/84248
27067         * config/i386/i386.c (ix86_option_override_internal): Mask out
27068         the CF_SET bit when checking -fcf-protection.
27070 2018-02-07  Tom de Vries  <tom@codesourcery.com>
27072         PR libgomp/84217
27073         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
27074         enough.
27076 2018-02-07  Richard Biener  <rguenther@suse.de>
27078         PR tree-optimization/84204
27079         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
27080         this place.
27082         PR tree-optimization/84205
27083         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
27084         special-case isl_ast_op_zdiv_r.
27086         PR tree-optimization/84223
27087         * graphite-scop-detection.c (gather_bbs::before_dom_children):
27088         Only add conditions from within the region.
27089         (gather_bbs::after_dom_children): Adjust.
27091 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
27093         PR target/84209
27094         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
27095         * config/avr/avr.md: Only post-reload split REG-REG moves if
27096         either register is GENERAL_REG_P.
27098 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
27100         PR tree-optimization/84235
27101         * tree-ssa-scopedtables.c
27102         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
27103         if the subtraction is performed in floating point type where NaNs are
27104         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
27105         build 1.  Formatting fix.
27107 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
27109         PR target/84146
27110         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
27111         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
27112         and skip it regardless of bb boundaries.  Use CALL_P macro,
27113         don't test INSN_P (insn) together with CALL_P or JUMP_P check
27114         unnecessarily, formatting fix.
27116 2018-02-06  Michael Collison  <michael.collison@arm.com>
27118         * config/arm/thumb2.md:
27119         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
27120         (*thumb_mov_notscc): Ditto.
27122 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
27124         PR target/84154
27125         * config/rs6000/rs6000.md (su code attribute): Use "u" for
27126         unsigned_fix, not "s".
27128 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27130         * configure.ac (gcc_fn_eh_frame_ro): New function.
27131         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
27132         correct .eh_frame permissions.
27133         * configure: Regenerate.
27135 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
27137         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
27138         irrelevant options.
27140 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27142         * config/rs6000/rs6000.c (rs6000_option_override_internal):
27143         Display warning message for -mno-speculate-indirect-jumps.
27145 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
27147         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
27148         Undocumented.
27149         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
27151 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
27153         PR tree-optimization/84225
27154         * tree-eh.c (find_trapping_overflow): Only call
27155         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
27157 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27159         PR target/84145
27160         * config/i386/i386.c: Reimplement the check of possible options
27161         -mibt/-mshstk conbination. Change error messages.
27162         * doc/invoke.texi: Fix a typo: remove extra '='.
27164 2018-02-06  Marek Polacek  <polacek@redhat.com>
27166         PR tree-optimization/84228
27167         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
27169 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
27171         PR target/82641
27172         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
27173         emitted arch directives.
27174         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
27175         __ARM_FEATURE_COPROC before changing architectures.
27177 2018-02-06  Richard Biener  <rguenther@suse.de>
27179         * config/i386/i386.c (print_reg): Fix typo.
27180         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
27182 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
27184         * configure: Regenerate.
27186 2018-02-05  Martin Sebor  <msebor@redhat.com>
27188         PR tree-optimization/83369
27189         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
27190         inlining context.
27192 2018-02-05  Martin Liska  <mliska@suse.cz>
27194         * doc/invoke.texi: Cherry-pick upstream r323995.
27196 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
27198         * ira.c (ira_init_register_move_cost): Adjust comment.
27200 2018-02-05  Martin Liska  <mliska@suse.cz>
27202         PR gcov-profile/84137
27203         * doc/gcov.texi: Fix typo in documentation.
27205 2018-02-05  Martin Liska  <mliska@suse.cz>
27207         PR gcov-profile/83879
27208         * doc/gcov.texi: Document necessity of --dynamic-list-data when
27209         using dlopen functionality.
27211 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
27213         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
27214         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
27215         _mm_maskz_range_ss, _mm_mask_range_round_ss,
27216         _mm_maskz_range_round_ss): New intrinsics.
27217         (__builtin_ia32_rangesd128_round)
27218         (__builtin_ia32_rangess128_round): Remove.
27219         (__builtin_ia32_rangesd128_mask_round,
27220         __builtin_ia32_rangess128_mask_round): New builtins.
27221         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
27222         __builtin_ia32_rangess128_round): Remove.
27223         (__builtin_ia32_rangesd128_mask_round,
27224         __builtin_ia32_rangess128_mask_round): New builtins.
27225         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
27226         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
27227         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
27228         "<round_saeonly_constraint>")): Changed to ...
27229         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
27230         "<round_saeonly_scalar_constraint>")): ... this.
27231         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
27232         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
27233         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
27234         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
27235         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
27237 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
27239         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
27240         options.
27241         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
27242         Remove all values except native, 8540 and 8548.
27244 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
27246         * config/i386/i386.c (ix86_output_function_return): Pass
27247         INVALID_REGNUM, instead of -1, as invalid register number to
27248         indirect_thunk_name and output_indirect_thunk.
27250 2018-02-02  Julia Koval  <julia.koval@intel.com>
27252         * config.gcc: Add -march=icelake.
27253         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
27254         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
27255         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
27256         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
27257         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
27258         (processor_target_table): Add icelake.
27259         (ix86_option_override_internal): Handle new PTAs.
27260         (get_builtin_code_for_version): Handle icelake.
27261         (M_INTEL_COREI7_ICELAKE): New.
27262         (fold_builtin_cpu): Handle icelake.
27263         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
27264         * doc/invoke.texi: Add -march=icelake.
27266 2018-02-02  Julia Koval  <julia.koval@intel.com>
27268         * config/i386/i386.c (ix86_option_override_internal): Change flags type
27269         to wide_int_bitmask.
27270         * wide-int-bitmask.h: New.
27272 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27274         PR target/84066
27275         * config/i386/i386.md: Replace Pmode with word_mode in
27276         builtin_setjmp_setup and builtin_longjmp to support x32.
27278 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
27280         PR target/56010
27281         PR target/83743
27282         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
27283         #include "opts.h".
27284         (rs6000_supported_cpu_names): New static variable.
27285         (linux_cpu_translation_table): Likewise.
27286         (elf_platform) <cpu>: Define new static variable and use it.
27287         Translate kernel AT_PLATFORM name to canonical name if needed.
27288         Error if platform name is unknown.
27290 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
27292         PR target/84089
27293         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
27295 2018-02-01  Jeff Law  <law@redhat.com>
27297         PR target/84128
27298         * config/i386/i386.c (release_scratch_register_on_entry): Add new
27299         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
27300         the scratch if RELEASE_VIA_POP is false.
27301         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
27302         If we have to save a temporary register, decrement SIZE appropriately.
27303         Pass new arguments to release_scratch_register_on_entry.
27304         (ix86_adjust_stack_and_probe): Likewise.
27305         (ix86_emit_probe_stack_range): Pass new arguments to
27306         release_scratch_register_on_entry.
27308 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
27310         PR rtl-optimization/84157
27311         * combine.c (change_zero_ext): Use REG_P predicate in
27312         front of HARD_REGISTER_P predicate.
27314 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
27316         * config/avr/avr.c (avr_option_override): Move disabling of
27317         -fdelete-null-pointer-checks to...
27318         * common/config/avr/avr-common.c (avr_option_optimization_table):
27319         ...here.
27321 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27323         PR tree-optimization/81635
27324         * tree-data-ref.c (split_constant_offset_1): For types that
27325         wrap on overflow, try to use range info to prove that wrapping
27326         cannot occur.
27328 2018-02-01  Renlin Li  <renlin.li@arm.com>
27330         PR target/83370
27331         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
27332         TAILCALL_ADDR_REGS.
27333         (aarch64_register_move_cost): Likewise.
27334         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
27335         TAILCALL_ADDR_REGS.
27336         (REG_CLASS_NAMES): Likewise.
27337         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
27338         TAILCALL_ADDR_REGS. Remove IP registers.
27339         * config/aarch64/aarch64.md (Ucs): Update register constraint.
27341 2018-02-01  Richard Biener  <rguenther@suse.de>
27343         * domwalk.h (dom_walker::dom_walker): Add additional constructor
27344         for specifying RPO order and allow NULL for that.
27345         * domwalk.c (dom_walker::dom_walker): Likewise.
27346         (dom_walker::walk): Handle NULL RPO order.
27347         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
27348         in RPO order.
27349         (rewrite_update_dom_walker): Likewise.
27350         (mark_def_dom_walker): Likewise.
27352 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27354         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
27355         (aarch64_maybe_expand_sve_subreg_move): Declare.
27356         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
27357         * config/aarch64/predicates.md (aarch64_any_register_operand): New
27358         predicate.
27359         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
27360         that are semantically a reverse operation.
27361         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
27362         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
27363         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
27364         functions.
27365         (aarch64_can_change_mode_class): For big-endian, forbid changes
27366         between two SVE modes if they have different element sizes.
27368 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27370         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
27371         the TImode handling for big-endian targets.
27373 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27375         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
27376         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
27377         not just bytes.
27378         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
27379         Remove BSWAP handing for big-endian targets and use the form of
27380         LD1RQ appropariate for the mode.
27382 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27384         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
27385         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
27386         duplicated element.
27388 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27390         PR tearget/83845
27391         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
27392         check for operands that need to go through aarch64_sve_reload_be.
27394 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
27396         PR tree-optimization/81661
27397         PR tree-optimization/84117
27398         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
27399         * tree-eh.c: Include gimplify.h.
27400         (find_trapping_overflow, replace_trapping_overflow,
27401         rewrite_to_non_trapping_overflow): New functions.
27402         * tree-vect-loop.c: Include tree-eh.h.
27403         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
27404         * tree-data-ref.c: Include tree-eh.h.
27405         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
27407 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
27409         PR rtl-optimization/84123
27410         * combine.c (change_zero_ext): Check if hard register satisfies
27411         can_change_dest_mode before calling gen_lowpart_SUBREG.
27413 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
27415         PR target/82444
27416         * ira.c (ira_init_register_move_cost): Remove assert.
27418 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27420         PR rtl-optimization/84071
27421         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
27422         * doc/tm.texi: Regenerate.
27424 2018-01-31  Richard Biener  <rguenther@suse.de>
27426         PR tree-optimization/84132
27427         * tree-data-ref.c (analyze_miv_subscript): Properly
27428         check whether evolution_function_is_affine_multivariate_p
27429         before calling gcd_of_steps_may_divide_p.
27431 2018-01-31  Julia Koval  <julia.koval@intel.com>
27433         PR target/83618
27434         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
27435         * config/i386/i386.md (rdpid_rex64) New.
27436         (rdpid): Make 32bit only.
27438 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
27440         PR lto/84105
27441         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
27442         an IDENTIFIER_NODE for FUNCTION_TYPE's.
27444 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27446         Revert
27447         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27449         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27451 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27453         PR rtl-optimization/84071
27454         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
27455         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
27457 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
27459         * config/arc/arc.c (arc_handle_aux_attribute): New function.
27460         (arc_attribute_table): Add 'aux' attribute.
27461         (arc_in_small_data_p): Consider aux like variables.
27462         (arc_is_aux_reg_p): New function.
27463         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
27464         (arc_get_aux_arg): New function.
27465         (prepare_move_operands): Handle aux-register access.
27466         (arc_handle_aux_attribute): New function.
27467         * doc/extend.texi (ARC Variable attributes): Add subsection.
27469 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
27471         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
27472         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
27473         (arc_attribute_table): Add 'uncached' attribute.
27474         (arc_print_operand): Print '.di' flag for uncached memory
27475         accesses.
27476         (arc_in_small_data_p): Do not consider for small data the uncached
27477         types.
27478         (arc_is_uncached_mem_p): New function.
27479         * config/arc/predicates.md (compact_store_memory_operand): Check
27480         for uncached memory accesses.
27481         (nonvol_nonimm_operand): Likewise.
27482         * doc/extend.texi (ARC Type Attribute): New subsection.
27484 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
27486         PR c/84100
27487         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
27488         falign-loops=): Add Optimization flag.
27490 2018-01-30  Jeff Law  <law@redhat.com>
27492         PR target/84064
27493         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
27494         INT_REGISTERS_SAVED.  Check it prior to calling
27495         get_scratch_register_on_entry.
27496         (ix86_adjust_stack_and_probe): Similarly.
27497         (ix86_emit_probe_stack_range): Similarly.
27498         (ix86_expand_prologue): Corresponding changes.
27500 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27502         PR target/40411
27503         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
27504         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
27506 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
27508         PR target/84112
27509         * lra-constraints.c (curr_insn_transform): Process AND in the
27510         address.
27512 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
27514         PR rtl-optimization/83986
27515         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
27516         dependence against last_pending_memory_flush in addition to
27517         pending_jump_insns.
27519 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
27521         PR tree-optimization/81611
27522         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
27523         copies.
27525 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27527         PR target/83758
27528         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
27529         a reg rtx.
27531 2018-01-30  Richard Biener  <rguenther@suse.de>
27532             Jakub Jelinek  <jakub@redhat.com>
27534         PR tree-optimization/84111
27535         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
27536         inner loops added during recursion, as they don't have up-to-date
27537         SSA form.
27539 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
27541         PR ipa/81360
27542         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
27543         (can_inline_edge_by_limits_p): ... here.
27544         (can_early_inline_edge_p, check_callers,
27545         update_caller_keys, update_callee_keys, recursive_inlining,
27546         add_new_edges_to_heap, speculation_useful_p,
27547         inline_small_functions,
27548         inline_small_functions, flatten_function,
27549         inline_to_all_callers_1): Update.
27551 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
27553         * profile-count.c (profile_count::combine_with_ipa_count): Handle
27554         zeros correctly.
27556 2018-01-30  Richard Biener  <rguenther@suse.de>
27558         PR tree-optimization/83008
27559         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
27560         invariant and constant vector uses in stmts when they need
27561         more than one stmt.
27563 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27565         PR bootstrap/84017
27566         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
27567         * configure: Regenerate.
27569 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
27571         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
27572         pattern.
27573         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
27574         Use gen_rtx_REG rather than gen_lowpart.
27576 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
27578         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
27579         rather than 0 when creating partial subregs.
27581 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
27583         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
27584         of usage.
27586 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
27588         PR target/81550
27589         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
27590         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
27591         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
27592         flags.  This restores the settings used before the 2017-07-24.
27593         Turning off pre increment/decrement/modify allows IVOPTS to
27594         optimize DF/SF loops where the index is an int.
27596 2018-01-29  Richard Biener  <rguenther@suse.de>
27597             Kelvin Nilsen  <kelvin@gcc.gnu.org>
27599         PR bootstrap/80867
27600         * tree-vect-stmts.c (vectorizable_call): Don't call
27601         targetm.vectorize_builtin_md_vectorized_function if callee is
27602         NULL.
27604 2018-01-22  Carl Love  <cel@us.ibm.com>
27606         * doc/extend.tex: Fix typo in second arg in
27607         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
27609 2018-01-29  Richard Biener  <rguenther@suse.de>
27611         PR tree-optimization/84086
27612         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
27613         (flush_ssaname_freelist): When SSA names were released reset
27614         the SCEV hash table.
27616 2018-01-29  Richard Biener  <rguenther@suse.de>
27618         PR tree-optimization/84057
27619         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
27620         removed paths when removing edges.
27622 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
27624         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
27625         -mfunction-return=@var{choice}.
27627 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27629         PR diagnostic/84034
27630         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
27631         Handle CR like TAB.
27632         (layout::print_source_line): Likewise.
27633         (test_get_line_width_without_trailing_whitespace): Add test cases.
27635 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
27637         PR middle-end/84040
27638         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
27639         debug insns.
27641 2018-01-26  Jim Wilson  <jimw@sifive.com>
27643         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
27645         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
27646         specified.
27648 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27650         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
27651         and CMP + SUB-immediate -> SUBS.
27653 2018-01-26  Martin Sebor  <msebor@redhat.com>
27655         PR tree-optimization/83896
27656         * tree-ssa-strlen.c (get_string_len): Rename...
27657         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
27658         Avoid assuming length is constant.
27659         (handle_char_store): Use HOST_WIDE_INT for string length.
27661 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
27663         PR target/81763
27664         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
27665         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
27667 2018-01-26  Richard Biener  <rguenther@suse.de>
27669         PR rtl-optimization/84003
27670         * dse.c (record_store): Only record redundant stores when
27671         the earlier store aliases at least all accesses the later one does.
27673 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
27675         PR rtl-optimization/83985
27676         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
27677         REG_CFA_RESTORE insns.
27678         (delete_unmarked_insns): Don't ignore separate shrink wrapping
27679         REG_CFA_RESTORE insns here.
27681         PR c/83989
27682         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
27683         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27685 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27687         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27688         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27689         (arc_init): Likewise.
27690         (arc_override_options): Likewise.
27691         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27692         value.
27693         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27694         support.
27695         * config/arc/arc.h (TARGET_DBNZ): Define.
27696         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27697         properly set the tune attribute.
27698         (dbnz): Use TARGET_DBNZ guard.
27699         * config/arc/arc.opt (mtune): Add core3 option.
27701 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27703         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27704         recognize new pic like addresses.
27705         (arc_delegitimize_address): Clean up.
27707 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27709         * config/arc/arc-arches.def: Option mrf16 valid for all
27710         architectures.
27711         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27712         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27713         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27714         * config/arc/arc-tables.opt: Regenerate.
27715         * config/arc/arc.c (arc_conditional_register_usage): Handle
27716         reduced register file case.
27717         (arc_file_start): Set must have build attributes.
27718         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27719         mrf16 option value.
27720         * config/arc/arc.opt (mrf16): Add new option.
27721         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27722         * config/arc/genmultilib.awk: Handle new mrf16 option.
27723         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27724         * config/arc/t-multilib: Regenerate.
27725         * doc/invoke.texi (ARC Options): Document mrf16 option.
27727 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27729         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27730         * config/arc/arc.c (arc_handle_secure_attribute): New function.
27731         (arc_attribute_table): Add 'secure_call' attribute.
27732         (arc_print_operand): Print secure call operand.
27733         (arc_function_ok_for_sibcall): Don't optimize tail calls when
27734         secure.
27735         (arc_is_secure_call_p): New function.  * config/arc/arc.md
27736         (call_i): Add support for sjli instruction.
27737         (call_value_i): Likewise.
27738         * config/arc/constraints.md (Csc): New constraint.
27740 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27741             John Eric Martin  <John.Martin@emmicro-us.com>
27743         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27744         * config/arc/arc.c (_arc_jli_section): New struct.
27745         (arc_jli_section): New type.
27746         (rc_jli_sections): New static variable.
27747         (arc_handle_jli_attribute): New function.
27748         (arc_attribute_table): Add jli_always and jli_fixed attribute.
27749         (arc_file_end): New function.
27750         (TARGET_ASM_FILE_END): Define.
27751         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27752         (arc_add_jli_section): New function.
27753         (jli_call_scan): Likewise.
27754         (arc_reorg): Call jli_call_scan.
27755         (arc_output_addsi): Remove 'S' from printing asm operand.
27756         (arc_is_jli_call_p): New function.
27757         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27758         operand.
27759         (movhi_insn): Likewise.
27760         (movsi_insn): Likewise.
27761         (movsi_set_cc_insn): Likewise.
27762         (loadqi_update): Likewise.
27763         (load_zeroextendqisi_update): Likewise.
27764         (load_signextendqisi_update): Likewise.
27765         (loadhi_update): Likewise.
27766         (load_zeroextendhisi_update): Likewise.
27767         (load_signextendhisi_update): Likewise.
27768         (loadsi_update): Likewise.
27769         (loadsf_update): Likewise.
27770         (movsicc_insn): Likewise.
27771         (bset_insn): Likewise.
27772         (bxor_insn): Likewise.
27773         (bclr_insn): Likewise.
27774         (bmsk_insn): Likewise.
27775         (bicsi3_insn): Likewise.
27776         (cmpsi_cc_c_insn): Likewise.
27777         (movsi_ne): Likewise.
27778         (movsi_cond_exec): Likewise.
27779         (clrsbsi2): Likewise.
27780         (norm_f): Likewise.
27781         (normw): Likewise.
27782         (swap): Likewise.
27783         (divaw): Likewise.
27784         (flag): Likewise.
27785         (sr): Likewise.
27786         (kflag): Likewise.
27787         (ffs): Likewise.
27788         (ffs_f): Likewise.
27789         (fls): Likewise.
27790         (call_i): Remove 'S' asm letter, add jli instruction.
27791         (call_value_i): Likewise.
27792         * config/arc/arc.op (mjli-always): New option.
27793         * config/arc/constraints.md (Cji): New constraint.
27794         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27795         operand.
27796         (subsf3_fpx): Likewise.
27797         (mulsf3_fpx): Likewise.
27798         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27799         asm operand.
27800         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27801         function attrbutes.
27802         * doc/invoke.texi (ARC): Document mjli-always option.
27804 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
27806         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27807         avoid addition with 0 and use incw and decw where possible.
27809 2018-01-26  Richard Biener  <rguenther@suse.de>
27811         PR tree-optimization/81082
27812         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27813         association if it requires casting to unsigned.
27814         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27815         from fold_plusminus_mult_expr to catch important cases late when
27816         range info is available.
27818 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27820         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27821         * configure.ac (hidden_linkonce): New test.
27822         * configure: Regenerate.
27823         * config.in: Regenerate.
27825 2018-01-26  Julia Koval  <julia.koval@intel.com>
27827         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27828         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27829         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27830         _mm_mask_bitshuffle_epi64_mask): Fix type.
27831         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27832         USI_FTYPE_V4DI_V4DI_USI): Remove.
27833         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27834         __builtin_ia32_vpshufbitqmb256_mask,
27835         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27836         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27837         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27839 2018-01-26  Alan Modra  <amodra@gmail.com>
27841         PR target/84033
27842         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27843         UNSPEC_VBPERMQ.  Sort other unspecs.
27845 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27847         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27849 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27851         PR middle-end/83055
27852         * predict.c (drop_profile): Do not push/pop cfun; update also
27853         node->count.
27854         (handle_missing_profiles): Fix logic looking for zero profiles.
27856 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27858         PR middle-end/83977
27859         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27860         on functions with #pragma omp declare simd or functions with simd
27861         attribute.
27862         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27863         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27864         Remove trailing \n from warning_at calls.
27866 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27868         PR target/84028
27869         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27870         for neutered workers.
27872 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27874         PR target/68467
27875         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27876         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27878 2018-01-24  Jeff Law  <law@redhat.com>
27880         PR target/83994
27881         * i386.c (get_probe_interval): Move to earlier point.
27882         (ix86_compute_frame_layout): If -fstack-clash-protection and
27883         the frame is larger than the probe interval, then use pushes
27884         to save registers rather than reg->mem moves.
27885         (ix86_expand_prologue): Remove conditional for int_registers_saved
27886         assertion.
27888 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27890         PR target/84014
27891         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27892         min/max for never referenced object.
27894 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27896         PR middle-end/83977
27897         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27898         here.
27899         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27900         attributes from DECL_ATTRIBUTES (decl) without affecting
27901         DECL_ATTRIBUTES (current_function_decl).
27902         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27903         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27905 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27907         PR tree-optimization/83979
27908         * fold-const.c (fold_comparison): Use constant_boolean_node
27909         instead of boolean_{true,false}_node.
27911 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27913         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27914         with zero counts.
27916 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27918         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27919         Simplify the clause that sets the length attribute.
27920         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27921         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27922         clause that sets the length attribute.
27923         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27925 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27927         PR target/83589
27928         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27929         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27930         Add strict parameter.
27931         (prevent_branch_around_nothing): Insert dummy insn between branch to
27932         label and label with no ptx insn inbetween.
27933         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27935 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27937         PR target/81352
27938         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27939         for neutered threads in warp.
27940         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27942 2018-01-24  Richard Biener  <rguenther@suse.de>
27944         PR tree-optimization/83176
27945         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27946         operands.
27948 2018-01-24  Richard Biener  <rguenther@suse.de>
27950         PR tree-optimization/82819
27951         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27952         code generating pluses that are no-ops in the target precision.
27954 2018-01-24  Richard Biener  <rguenther@suse.de>
27956         PR middle-end/84000
27957         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27959 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27961         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27962         to merge probabilities.
27963         * predict.c (probably_never_executed): Also mark as cold functions
27964         with global 0 profile and guessed local profile.
27965         * profile-count.c (profile_probability::combine_with_count): New
27966         member function.
27967         * profile-count.h (profile_probability::operator*,
27968         profile_probability::operator*=, profile_probability::operator/,
27969         profile_probability::operator/=): Reduce precision to adjusted
27970         and set value to guessed on contradictory divisions.
27971         (profile_probability::combine_with_freq): Remove.
27972         (profile_probability::combine_wiht_count): Declare.
27973         (profile_count::force_nonzero):: Set to adjusted.
27974         (profile_count::probability_in):: Set quality to adjusted.
27975         * tree-ssa-tail-merge.c (replace_block_by): Use
27976         combine_with_count.
27978 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27979             Jim Wilson  <jimw@sifive.com>
27981         * config/riscv/riscv.c (riscv_stack_boundary): New.
27982         (riscv_option_override): Set riscv_stack_boundary.  Handle
27983         riscv_preferred_stack_boundary_arg.
27984         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27985         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27986         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27987         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27988         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27989         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27991 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27993         PR target/83905
27994         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27995         of struct ix86_frame.
27996         (ix86_expand_epilogue): Likewise.  Add a local variable for
27997         the reg_save_offset field in struct ix86_frame.
27999 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
28001         PR tree-optimization/82604
28002         * tree-loop-distribution.c (enum partition_kind): New enum item
28003         PKIND_PARTIAL_MEMSET.
28004         (partition_builtin_p): Support above new enum item.
28005         (generate_code_for_partition): Ditto.
28006         (compute_access_range): Differentiate cases that equality can be
28007         proven at all loops, the innermost loops or no loops.
28008         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
28009         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
28010         (finalize_partitions, distribute_loop): Don't fuse partition of
28011         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
28012         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
28013         parloop is enabled.
28015 2018-01-23  Martin Liska  <mliska@suse.cz>
28017         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
28018         order to ignore the predictor.
28019         (PRED_POLYMORPHIC_CALL): Likewise.
28020         (PRED_RECURSIVE_CALL): Likewise.
28022 2018-01-23  Martin Liska  <mliska@suse.cz>
28024         * tree-profile.c (tree_profiling): Print function header to
28025         aware reader which function we are working on.
28026         * value-prof.c (gimple_find_values_to_profile): Do not print
28027         not interesting value histograms.
28029 2018-01-23  Martin Liska  <mliska@suse.cz>
28031         * profile-count.h (enum profile_quality): Add
28032         profile_uninitialized as the first value. Do not number values
28033         as they are zero based.
28034         (profile_count::verify): Update sanity check.
28035         (profile_probability::verify): Likewise.
28037 2018-01-23  Nathan Sidwell  <nathan@acm.org>
28039         * doc/invoke.texi (ffor-scope): Deprecate.
28041 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28043         PR tree-optimization/83510
28044         * domwalk.c (set_all_edges_as_executable): New function.
28045         (dom_walker::dom_walker): Convert bool param
28046         "skip_unreachable_blocks" to enum reachability.  Move setup of
28047         edge flags to set_all_edges_as_executable and only do it when
28048         reachability is REACHABLE_BLOCKS.
28049         * domwalk.h (enum dom_walker::reachability): New enum.
28050         (dom_walker::dom_walker): Convert bool param
28051         "skip_unreachable_blocks" to enum reachability.
28052         (set_all_edges_as_executable): New decl.
28053         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
28054         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
28055         "reachability".
28056         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
28057         but converting true to REACHABLE_BLOCKS.
28058         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
28059         * tree-vrp.c
28060         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
28061         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
28062         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
28063         REACHABLE_BLOCKS.
28064         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
28065         if check_all_array_refs will be called.
28067 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28069         * tree.c (selftest::test_location_wrappers): Add more test
28070         coverage.
28072 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28074         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
28075         (selftest::test_bit_in_range): Likewise.
28077 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
28079         PR testsuite/83888
28080         * doc/sourcebuild.texi (vect_float): Say that the selector
28081         only describes the situation when -funsafe-math-optimizations is on.
28082         (vect_float_strict): Document.
28084 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
28086         PR tree-optimization/83965
28087         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
28088         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
28089         instead of checking only for a reduction.
28090         (vect_recog_widen_sum_pattern): Likewise.
28092 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
28094         * predict.c (probably_never_executed): Only use precise profile info.
28095         (compute_function_frequency): Skip after inlining hack since we now
28096         have quality checking.
28098 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
28100         * profile-count.h (profile_probability::very_unlikely,
28101         profile_probability::unlikely, profile_probability::even): Set
28102         precision to guessed.
28104 2018-01-23  Richard Biener  <rguenther@suse.de>
28106         PR tree-optimization/83963
28107         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
28108         Properly terminate dominator walk when crossing the exit edge not
28109         when visiting its source block.
28111 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
28113         PR c++/83918
28114         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
28115         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
28117 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
28119         PR tree-optimization/83957
28120         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
28121         semicolon after for body surrounded by braces.
28123         PR tree-optimization/83081
28124         * profile-count.h (profile_probability::split): New method.
28125         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
28126         Use profile_probability::split.
28127         (do_compare_rtx_and_jump): Fix adjustment of probabilities
28128         when splitting a single conditional jump into 2.
28130 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
28132         PR tree-optimization/69452
28133         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
28134         decl.
28136 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28138         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
28139         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
28140         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
28142 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28144         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
28145         declaration.
28146         * config/rl78/rl78.md (movdi): New define_expand.
28147         * config/rl78/rl78.c (rl78_split_movdi): New function.
28149 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
28151         PR target/83862
28152         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
28153         no longer used.
28154         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
28155         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
28156         128-bit to produce an UNSPEC move to get the double word with the
28157         signbit and then a shift directly to do signbit.
28158         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
28159         implementation with a new version that just does either a direct
28160         move or a regular move.  Move memory interface to separate insns.
28161         Move insns so they are next to the expander.
28162         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
28163         with signbit move.  Split big and little endian case.
28164         (signbit<mode>2_dm_mem_le): Likewise.
28165         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
28166         (signbit<mode>2_dm2): Likewise.
28168 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28170         * config/rl78/rl78.md (anddi3): New define_expand.
28172 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28174         * config/rl78/rl78.md (umindi3): New define_expand.
28176 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28178         * config/rl78/rl78.md (smindi3): New define_expand.
28180 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28182         * config/rl78/rl78.md (smaxdi3): New define_expand.
28184 2018-01-22  Carl Love  <cel@us.ibm.com>
28186         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
28187         LVX_V1TI): Add macro expansion.
28188         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
28189         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
28190         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
28191         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
28192         Change check to determine if the instruction is a byte reversing
28193         entry.  Fix typo in comment.
28194         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
28195         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
28196         Add def_builtin calls for new builtins.
28197         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
28198         Add define_insn expansion.
28200 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28202         * config/rl78/rl78.md (umaxdi3): New define_expand.
28204 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28206         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
28207         for non-QImode registers.
28209 2018-01-22  Richard Biener  <rguenther@suse.de>
28211         PR tree-optimization/83963
28212         * graphite-scop-detection.c (scop_detection::get_sese): Delay
28213         including the loop exit block.
28214         (scop_detection::merge_sese): Likewise.
28215         (scop_detection::add_scop): Do it here instead.
28217 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28219         * doc/sourcebuild.texi (arm_softfloat): Document.
28221 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
28223         PR gcc/77734
28224         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
28225         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
28226         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
28228 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28229             David Edelsohn  <dje.gcc@gmail.com>
28231         PR target/83946
28232         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28233         Change "crset eq" to "crset 2".
28234         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28235         (*call_indirect_aix<mode>_nospec): Likewise.
28236         (*call_value_indirect_aix<mode>_nospec): Likewise.
28237         (*call_indirect_elfv2<mode>_nospec): Likewise.
28238         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
28239         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
28240         change assembly output from . to $.
28241         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28242         (indirect_jump<mode>_nospec): Change assembly output from . to $.
28243         (*tablejump<mode>_internal1_nospec): Likewise.
28245 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
28247         PR target/80870
28248         * config/sh/sh_optimize_sett_clrt.cc:
28249         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
28251 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
28253         PR tree-optimization/83940
28254         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
28255         offset_dt to vect_constant_def rather than vect_unknown_def_type.
28256         (vect_check_load_store_mask): Add a mask_dt_out parameter and
28257         use it to pass back the definition type.
28258         (vect_check_store_rhs): Likewise rhs_dt_out.
28259         (vect_build_gather_load_calls): Add a mask_dt argument and use
28260         it instead of a call to vect_is_simple_use.
28261         (vectorizable_store): Update calls to vect_check_load_store_mask
28262         and vect_check_store_rhs.  Use the dt returned by the latter instead
28263         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
28264         instead of calls to vect_is_simple_use.  Pass the scalar rather
28265         than the vector operand to vect_is_simple_use when handling
28266         second and subsequent copies of an rhs value.
28267         (vectorizable_load): Update calls to vect_check_load_store_mask
28268         and vect_build_gather_load_calls.  Use the cached mask_dt and
28269         gs_info.offset_dt instead of calls to vect_is_simple_use.
28271 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
28273         PR middle-end/83945
28274         * tree-emutls.c: Include gimplify.h.
28275         (lower_emutls_2): New function.
28276         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
28277         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
28278         it before further processing.
28280         PR target/83930
28281         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
28282         UINTVAL (trueop1) instead of INTVAL (op1).
28284 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
28286         PR debug/81570
28287         PR debug/83728
28288         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
28289         INCOMING_FRAME_SP_OFFSET if not defined.
28290         (scan_trace): Add ENTRY argument.  If true and
28291         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
28292         emit a note to adjust the CFA offset.
28293         (create_cfi_notes): Adjust scan_trace callers.
28294         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
28295         INCOMING_FRAME_SP_OFFSET in the CIE.
28296         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
28297         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
28298         Likewise.
28299         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
28300         * doc/tm.texi: Regenerated.
28302 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28304         PR rtl-optimization/83147
28305         * lra-constraints.c (remove_inheritance_pseudos): Use
28306         lra_substitute_pseudo_within_insn.
28308 2018-01-19  Tom de Vries  <tom@codesourcery.com>
28309             Cesar Philippidis  <cesar@codesourcery.com>
28311         PR target/83920
28312         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
28314 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
28316         PR target/83790
28317         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
28318         spaces for function labels.
28320 2018-01-19  Martin Liska  <mliska@suse.cz>
28322         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
28323         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
28324         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
28325         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
28326         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
28327         (PRED_CONST_RETURN): Change from 69 to 65.
28328         (PRED_NULL_RETURN): Change from 91 to 71.
28329         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
28330         (PRED_LOOP_GUARD): Change from 66 to 73.
28332 2018-01-19  Martin Liska  <mliska@suse.cz>
28334         * predict.c (predict_insn_def): Add new assert.
28335         (struct branch_predictor): Change type to signed integer.
28336         (test_prediction_value_range): Amend test to cover
28337         PROB_UNINITIALIZED.
28338         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
28339         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
28340         (PRED_LOOP_ITERATIONS_MAX): Likewise.
28341         (PRED_LOOP_IV_COMPARE): Likewise.
28342         * predict.h (PROB_UNINITIALIZED): Define new constant.
28344 2018-01-19  Martin Liska  <mliska@suse.cz>
28346         * predict.c (dump_prediction): Add new format for
28347         analyze_brprob.py script which is enabled with -details
28348         suboption.
28349         * profile-count.h (precise_p): New function.
28351 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
28353         PR tree-optimization/83922
28354         * tree-vect-loop.c (vect_verify_full_masking): Return false if
28355         there are no statements that need masking.
28356         (vect_active_double_reduction_p): New function.
28357         (vect_analyze_loop_operations): Use it when handling phis that
28358         are not in the loop header.
28360 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
28362         PR tree-optimization/83914
28363         * tree-vect-loop.c (vectorizable_induction): Don't convert
28364         init_expr or apply the peeling adjustment for inductions
28365         that are nested within the vectorized loop.
28367 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28369         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
28370         instead of NEG.
28372 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
28374         PR sanitizer/81715
28375         PR testsuite/83882
28376         * function.h (gimplify_parameters): Add gimple_seq * argument.
28377         * function.c: Include gimple.h and options.h.
28378         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
28379         for the added local temporaries if needed.
28380         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
28381         if there are any parameter cleanups, wrap whole body into a
28382         try/finally with the cleanups.
28384 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
28386         PR target/82964
28387         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
28388         Use GET_MODE_CLASS for scalar floating point.
28390 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
28392         PR ipa/82256
28393         patch by PaX Team
28394         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
28395         Fix call of call_cgraph_insertion_hooks.
28397 2018-01-18  Martin Sebor  <msebor@redhat.com>
28399         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
28401 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
28403         PR ipa/83619
28404         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
28405         frequencies.
28407 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
28409         PR other/70268
28410         * common.opt: (-ffile-prefix-map): New option.
28411         * opts.c (common_handle_option): Defer it.
28412         * opts-global.c (handle_common_deferred_options): Handle it.
28413         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
28414         * file-prefix-map.h: New file.
28415         (remap_debug_filename, add_debug_prefix_map): ...here.
28416         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
28417         * final.c (debug_prefix_map, add_debug_prefix_map
28418         remap_debug_filename): Move to...
28419         * file-prefix-map.c: New file.
28420         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
28421         generalize, get rid of alloca(), use strrchr() instead of strchr().
28422         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
28423         Implement in terms of add_prefix_map().
28424         (remap_macro_filename, remap_debug_filename): Implement in term of
28425         remap_filename().
28426         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
28427         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
28428         * dbxout.c: Include file-prefix-map.h.
28429         * varasm.c: Likewise.
28430         * vmsdbgout.c: Likewise.
28431         * xcoffout.c: Likewise.
28432         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
28433         * doc/cppopts.texi (-fmacro-prefix-map): Document.
28434         * doc/invoke.texi (-ffile-prefix-map): Document.
28435         (-fdebug-prefix-map): Update description.
28437 2018-01-18  Martin Liska  <mliska@suse.cz>
28439         * config/i386/i386.c (indirect_thunk_name): Document that also
28440         lfence is emitted.
28441         (output_indirect_thunk): Document why both instructions
28442         (pause and lfence) are generated.
28444 2018-01-18  Richard Biener  <rguenther@suse.de>
28446         PR tree-optimization/83887
28447         * graphite-scop-detection.c
28448         (scop_detection::get_nearest_dom_with_single_entry): Remove.
28449         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
28450         (scop_detection::merge_sese): Re-implement with a flood-fill
28451         algorithm that properly finds a SESE region if it exists.
28453 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
28455         PR c/61240
28456         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
28457         pointer_diff optimizations use view_convert instead of convert.
28459 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28461         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28462         Generate different code for -mno-speculate-indirect-jumps.
28463         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28464         (*call_indirect_aix<mode>): Disable for
28465         -mno-speculate-indirect-jumps.
28466         (*call_indirect_aix<mode>_nospec): New define_insn.
28467         (*call_value_indirect_aix<mode>): Disable for
28468         -mno-speculate-indirect-jumps.
28469         (*call_value_indirect_aix<mode>_nospec): New define_insn.
28470         (*sibcall_nonlocal_sysv<mode>): Generate different code for
28471         -mno-speculate-indirect-jumps.
28472         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28474 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
28476         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
28477         long double type, set the flags for noting the default long double
28478         type, even if we don't pass or return a long double type.
28480 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
28482         PR ipa/83051
28483         * ipa-inline.c (flatten_function): Do not overwrite final inlining
28484         failure.
28486 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
28488         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
28489         support for merge[hl].
28490         (fold_mergehl_helper): New helper function.
28491         (tree-vector-builder.h): New #include for tree_vector_builder usage.
28492         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
28493         (altivec_vmrglw_direct): Add xxmrglw insn.
28495 2018-01-17  Andrew Waterman  <andrew@sifive.com>
28497         * config/riscv/riscv.c (riscv_conditional_register_usage): If
28498         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
28500 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
28502         PR lto/83121
28503         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
28504         call the lto_location_cache before reading the
28505         DECL_SOURCE_LOCATION of the types.
28507 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
28508             Richard Sandiford  <richard.sandiford@linaro.org>
28510         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
28511         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
28512         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
28513         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
28514         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
28515         Add declaration.
28516         * config/aarch64/constraints.md (aarch64_movti_operand):
28517         Limit immediates.
28518         * config/aarch64/predicates.md (Uti): Add new constraint.
28520 2018-01-17  Carl Love  <cel@us.ibm.com>
28522         * config/rs6000/vsx.md (define_expand xl_len_r,
28523         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
28524         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
28525         lxvll.
28526         (define_expand, define_insn): Move the shift left from  the
28527         define_insn to the define_expand for lxvl and stxvl instructions.
28528         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
28529         and XL_LEN_R definitions to PURE.
28531 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
28533         * config/i386/i386.c (indirect_thunk_name): Declare regno
28534         as unsigned int.  Compare regno with INVALID_REGNUM.
28535         (output_indirect_thunk): Ditto.
28536         (output_indirect_thunk_function): Ditto.
28537         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
28538         in the call to output_indirect_thunk_function.
28540 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
28542         PR middle-end/83884
28543         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
28544         rather than the size of inner_type to determine the stack slot size
28545         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
28547 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
28549         PR target/83546
28550         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
28551         to PTA_SILVERMONT.
28553 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
28555         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
28556         endian Linux systems to optionally enable multilibs for selecting
28557         the long double type if the user configured an explicit type.
28558         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
28559         have no long double multilibs if not defined.
28560         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
28561         warn if the user used -mabi={ieee,ibm}longdouble and we built
28562         multilibs for long double.
28563         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
28564         appropriate multilib option.
28565         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
28566         multilib options.
28567         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
28568         for building long double multilibs.
28569         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
28571 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
28573         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
28574         copies.
28576         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
28577         64 bits.
28578         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
28579         128 bits.
28581         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
28582         variables.
28584         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
28585         return value.
28587 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
28589         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
28590         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
28592 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28594         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
28595         different rtl trees depending on TARGET_64BIT.
28596         (rs6000_gen_lvx): Likewise.
28598 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
28600         * config/visium/visium.md (nop): Tweak comment.
28601         (hazard_nop): Likewise.
28603 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28605         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
28606         -mspeculate-indirect-jumps.
28607         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
28608         for -mno-speculate-indirect-jumps.
28609         (*call_indirect_elfv2<mode>_nospec): New define_insn.
28610         (*call_value_indirect_elfv2<mode>): Disable for
28611         -mno-speculate-indirect-jumps.
28612         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
28613         (indirect_jump): Emit different RTL for
28614         -mno-speculate-indirect-jumps.
28615         (*indirect_jump<mode>): Disable for
28616         -mno-speculate-indirect-jumps.
28617         (*indirect_jump<mode>_nospec): New define_insn.
28618         (tablejump): Emit different RTL for
28619         -mno-speculate-indirect-jumps.
28620         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
28621         (tablejumpsi_nospec): New define_expand.
28622         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
28623         (tablejumpdi_nospec): New define_expand.
28624         (*tablejump<mode>_internal1): Disable for
28625         -mno-speculate-indirect-jumps.
28626         (*tablejump<mode>_internal1_nospec): New define_insn.
28627         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
28628         option.
28630 2018-01-16  Artyom Skrobov tyomitch@gmail.com
28632         * caller-save.c (insert_save): Drop unnecessary parameter.  All
28633         callers updated.
28635 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28636             Richard Biener  <rguenth@suse.de>
28638         PR libgomp/83590
28639         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
28640         return early, inline manually is_gimple_sizepos.  Make sure if we
28641         call gimplify_expr we don't end up with a gimple constant.
28642         * tree.c (variably_modified_type_p): Don't return true for
28643         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
28644         * gimplify.h (is_gimple_sizepos): Remove.
28646 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28648         PR tree-optimization/83857
28649         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
28650         vectorizable_live_operation for pure SLP statements.
28651         (vectorizable_live_operation): Handle PHIs.
28653 2018-01-16  Richard Biener  <rguenther@suse.de>
28655         PR tree-optimization/83867
28656         * tree-vect-stmts.c (vect_transform_stmt): Precompute
28657         nested_in_vect_loop_p since the scalar stmt may get invalidated.
28659 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28661         PR c/83844
28662         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
28663         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
28664         If off is not INTEGER_CST, issue a may not be aligned warning
28665         rather than isn't aligned.  Use isn%'t rather than isn't.
28666         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
28667         into MULT_EXPR.
28668         <case MULT_EXPR>: Improve the case when bottom and one of the
28669         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
28670         operand, in that case check if the other operand is multiple of
28671         bottom divided by the INTEGER_CST operand.
28673 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28675         PR target/83858
28676         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
28677         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
28678         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
28679         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
28680         * config/pa/pa.c (pa_function_arg_advance): Likewise.
28681         (pa_function_arg, pa_arg_partial_bytes): Likewise.
28682         (pa_function_arg_size): New function.
28684 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28686         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28687         in a separate statement.
28689 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28691         PR tree-optimization/83847
28692         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28693         group gathers and scatters.
28695 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28697         PR rtl-optimization/86620
28698         * params.def (max-sched-ready-insns): Bump minimum value to 1.
28700         PR rtl-optimization/83213
28701         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28702         to last if both are JUMP_INSNs.
28704         PR tree-optimization/83843
28705         * gimple-ssa-store-merging.c
28706         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28707         store_immediate_info for bswap/nop orig_stores.
28709 2018-01-15  Andrew Waterman  <andrew@sifive.com>
28711         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28712         !TARGET_MUL.
28713         <UDIV>: Increase cost if !TARGET_DIV.
28715 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
28717         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28718         (define_attr "cr_logical_3op"): New.
28719         (cceq_ior_compare): Adjust.
28720         (cceq_ior_compare_complement): Adjust.
28721         (*cceq_rev_compare): Adjust.
28722         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28723         (is_cracked_insn): Adjust.
28724         (insn_must_be_first_in_group): Adjust.
28725         * config/rs6000/40x.md: Adjust.
28726         * config/rs6000/440.md: Adjust.
28727         * config/rs6000/476.md: Adjust.
28728         * config/rs6000/601.md: Adjust.
28729         * config/rs6000/603.md: Adjust.
28730         * config/rs6000/6xx.md: Adjust.
28731         * config/rs6000/7450.md: Adjust.
28732         * config/rs6000/7xx.md: Adjust.
28733         * config/rs6000/8540.md: Adjust.
28734         * config/rs6000/cell.md: Adjust.
28735         * config/rs6000/e300c2c3.md: Adjust.
28736         * config/rs6000/e500mc.md: Adjust.
28737         * config/rs6000/e500mc64.md: Adjust.
28738         * config/rs6000/e5500.md: Adjust.
28739         * config/rs6000/e6500.md: Adjust.
28740         * config/rs6000/mpc.md: Adjust.
28741         * config/rs6000/power4.md: Adjust.
28742         * config/rs6000/power5.md: Adjust.
28743         * config/rs6000/power6.md: Adjust.
28744         * config/rs6000/power7.md: Adjust.
28745         * config/rs6000/power8.md: Adjust.
28746         * config/rs6000/power9.md: Adjust.
28747         * config/rs6000/rs64.md: Adjust.
28748         * config/rs6000/titan.md: Adjust.
28750 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28752         * config/i386/predicates.md (indirect_branch_operand): Rewrite
28753         ix86_indirect_branch_register logic.
28755 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28757         * config/i386/constraints.md (Bs): Update
28758         ix86_indirect_branch_register check.  Don't check
28759         ix86_indirect_branch_register with GOT_memory_operand.
28760         (Bw): Likewise.
28761         * config/i386/predicates.md (GOT_memory_operand): Don't check
28762         ix86_indirect_branch_register here.
28763         (GOT32_symbol_operand): Likewise.
28765 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28767         * config/i386/predicates.md (constant_call_address_operand):
28768         Rewrite ix86_indirect_branch_register logic.
28769         (sibcall_insn_operand): Likewise.
28771 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28773         * config/i386/constraints.md (Bs): Replace
28774         ix86_indirect_branch_thunk_register with
28775         ix86_indirect_branch_register.
28776         (Bw): Likewise.
28777         * config/i386/i386.md (indirect_jump): Likewise.
28778         (tablejump): Likewise.
28779         (*sibcall_memory): Likewise.
28780         (*sibcall_value_memory): Likewise.
28781         Peepholes of indirect call and jump via memory: Likewise.
28782         * config/i386/i386.opt: Likewise.
28783         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28784         (GOT_memory_operand): Likewise.
28785         (call_insn_operand): Likewise.
28786         (sibcall_insn_operand): Likewise.
28787         (GOT32_symbol_operand): Likewise.
28789 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28791         PR middle-end/83837
28792         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28793         type rather than type addr's type points to.
28794         (expand_omp_atomic_mutex): Likewise.
28795         (expand_omp_atomic): Likewise.
28797 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28799         PR target/83839
28800         * config/i386/i386.c (output_indirect_thunk_function): Use
28801         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28802         for  __x86_return_thunk.
28804 2018-01-15  Richard Biener  <rguenther@suse.de>
28806         PR middle-end/83850
28807         * expmed.c (extract_bit_field_1): Fix typo.
28809 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28811         PR target/83687
28812         * config/arm/iterators.md (VF): New mode iterator.
28813         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28814         Remove integer-related logic from pattern.
28815         (neon_vabd<mode>_3): Likewise.
28817 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28819         PR middle-end/82694
28820         * common.opt (fstrict-overflow): No longer an alias.
28821         (fwrapv-pointer): New option.
28822         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28823         also for pointer types based on flag_wrapv_pointer.
28824         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28825         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28826         opts->x_flag_wrapv got set.
28827         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28828         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28829         POINTER_TYPE_OVERFLOW_UNDEFINED.
28830         * match.pd: Likewise in address comparison pattern.
28831         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28833 2018-01-15  Richard Biener  <rguenther@suse.de>
28835         PR lto/83804
28836         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28837         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28838         Reset type names to their identifier if their TYPE_DECL doesn't
28839         have linkage (and thus is used for ODR and devirt).
28840         (save_debug_info_for_decl): Remove.
28841         (save_debug_info_for_type): Likewise.
28842         (add_tree_to_fld_list): Adjust.
28843         * tree-pretty-print.c (dump_generic_node): Make dumping of
28844         type names more robust.
28846 2018-01-15  Richard Biener  <rguenther@suse.de>
28848         * BASE-VER: Bump to 8.0.1.
28850 2018-01-14  Martin Sebor  <msebor@redhat.com>
28852         PR other/83508
28853         * builtins.c (check_access): Avoid warning when the no-warning bit
28854         is set.
28856 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28858         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28859         * ira-color (allocno_hard_regs_compare): Likewise.
28861 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28863         PR target/83013
28864         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28865         Use .pushsection/.popsection.
28867 2018-01-14  Martin Sebor  <msebor@redhat.com>
28869         PR c++/81327
28870         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28872 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28874         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28875         entry from extra_headers.
28876         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28877         extra_headers, make the list bitwise identical to the i?86-*-* one.
28879 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28881         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28882         -mcmodel=large with -mindirect-branch=thunk,
28883         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28884         -mfunction-return=thunk-extern.
28885         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28886         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28887         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28889 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28891         * config/i386/i386.c (print_reg): Print the name of the full
28892         integer register without '%'.
28893         (ix86_print_operand): Handle 'V'.
28894         * doc/extend.texi: Document 'V' modifier.
28896 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28898         * config/i386/constraints.md (Bs): Disallow memory operand for
28899         -mindirect-branch-register.
28900         (Bw): Likewise.
28901         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28902         (GOT_memory_operand): Likewise.
28903         (call_insn_operand): Likewise.
28904         (sibcall_insn_operand): Likewise.
28905         (GOT32_symbol_operand): Likewise.
28906         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28907         for -mindirect-branch-register.
28908         (tablejump): Likewise.
28909         (*sibcall_memory): Likewise.
28910         (*sibcall_value_memory): Likewise.
28911         Disallow peepholes of indirect call and jump via memory for
28912         -mindirect-branch-register.
28913         (*call_pop): Replace m with Bw.
28914         (*call_value_pop): Likewise.
28915         (*sibcall_pop_memory): Replace m with Bs.
28916         * config/i386/i386.opt (mindirect-branch-register): New option.
28917         * doc/invoke.texi: Document -mindirect-branch-register option.
28919 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28921         * config/i386/i386-protos.h (ix86_output_function_return): New.
28922         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28923         set function_return_type.
28924         (indirect_thunk_name): Add ret_p to indicate thunk for function
28925         return.
28926         (output_indirect_thunk_function): Pass false to
28927         indirect_thunk_name.
28928         (ix86_output_indirect_branch_via_reg): Likewise.
28929         (ix86_output_indirect_branch_via_push): Likewise.
28930         (output_indirect_thunk_function): Create alias for function
28931         return thunk if regno < 0.
28932         (ix86_output_function_return): New function.
28933         (ix86_handle_fndecl_attribute): Handle function_return.
28934         (ix86_attribute_table): Add function_return.
28935         * config/i386/i386.h (machine_function): Add
28936         function_return_type.
28937         * config/i386/i386.md (simple_return_internal): Use
28938         ix86_output_function_return.
28939         (simple_return_internal_long): Likewise.
28940         * config/i386/i386.opt (mfunction-return=): New option.
28941         (indirect_branch): Mention -mfunction-return=.
28942         * doc/extend.texi: Document function_return function attribute.
28943         * doc/invoke.texi: Document -mfunction-return= option.
28945 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28947         * config/i386/i386-opts.h (indirect_branch): New.
28948         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28949         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28950         with local indirect jump when converting indirect call and jump.
28951         (ix86_set_indirect_branch_type): New.
28952         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28953         (indirectlabelno): New.
28954         (indirect_thunk_needed): Likewise.
28955         (indirect_thunk_bnd_needed): Likewise.
28956         (indirect_thunks_used): Likewise.
28957         (indirect_thunks_bnd_used): Likewise.
28958         (INDIRECT_LABEL): Likewise.
28959         (indirect_thunk_name): Likewise.
28960         (output_indirect_thunk): Likewise.
28961         (output_indirect_thunk_function): Likewise.
28962         (ix86_output_indirect_branch_via_reg): Likewise.
28963         (ix86_output_indirect_branch_via_push): Likewise.
28964         (ix86_output_indirect_branch): Likewise.
28965         (ix86_output_indirect_jmp): Likewise.
28966         (ix86_code_end): Call output_indirect_thunk_function if needed.
28967         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28968         needed.
28969         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28970         (ix86_attribute_table): Add indirect_branch.
28971         * config/i386/i386.h (machine_function): Add indirect_branch_type
28972         and has_local_indirect_jump.
28973         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28974         to true.
28975         (tablejump): Likewise.
28976         (*indirect_jump): Use ix86_output_indirect_jmp.
28977         (*tablejump_1): Likewise.
28978         (simple_return_indirect_internal): Likewise.
28979         * config/i386/i386.opt (mindirect-branch=): New option.
28980         (indirect_branch): New.
28981         (keep): Likewise.
28982         (thunk): Likewise.
28983         (thunk-inline): Likewise.
28984         (thunk-extern): Likewise.
28985         * doc/extend.texi: Document indirect_branch function attribute.
28986         * doc/invoke.texi: Document -mindirect-branch= option.
28988 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28990         PR ipa/83051
28991         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28993 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28995         * ipa-inline.c (want_inline_small_function_p): Return false if
28996         inlining has already failed with CIF_FINAL_ERROR.
28997         (update_caller_keys): Call want_inline_small_function_p before
28998         can_inline_edge_p.
28999         (update_callee_keys): Likewise.
29001 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
29003         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
29004         New function.
29005         (rs6000_quadword_masked_address_p): Likewise.
29006         (quad_aligned_load_p): Likewise.
29007         (quad_aligned_store_p): Likewise.
29008         (const_load_sequence_p): Add comment to describe the outer-most loop.
29009         (mimic_memory_attributes_and_flags): New function.
29010         (rs6000_gen_stvx): Likewise.
29011         (replace_swapped_aligned_store): Likewise.
29012         (rs6000_gen_lvx): Likewise.
29013         (replace_swapped_aligned_load): Likewise.
29014         (replace_swapped_load_constant): Capitalize argument name in
29015         comment describing this function.
29016         (rs6000_analyze_swaps): Add a third pass to search for vector loads
29017         and stores that access quad-word aligned addresses and replace
29018         with stvx or lvx instructions when appropriate.
29019         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
29020         New function prototype.
29021         (rs6000_quadword_masked_address_p): Likewise.
29022         (rs6000_gen_lvx): Likewise.
29023         (rs6000_gen_stvx): Likewise.
29024         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
29025         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
29026         when memory address is aligned.
29027         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
29028         this split to select lvx instruction when memory address is aligned.
29029         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
29030         instruction when memory address is aligned.
29031         (*vsx_le_perm_load_v16qi): Likewise.
29032         (four unnamed splitters): Modify to select the stvx instruction
29033         when memory is aligned.
29035 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
29037         * predict.c (determine_unlikely_bbs): Handle correctly BBs
29038         which appears in the queue multiple times.
29040 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29041             Alan Hayward  <alan.hayward@arm.com>
29042             David Sherwood  <david.sherwood@arm.com>
29044         * tree-vectorizer.h (vec_lower_bound): New structure.
29045         (_loop_vec_info): Add check_nonzero and lower_bounds.
29046         (LOOP_VINFO_CHECK_NONZERO): New macro.
29047         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
29048         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
29049         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
29050         fields.  Make seg_len the distance travelled, not including the
29051         access size.
29052         (dr_direction_indicator): Declare.
29053         (dr_zero_step_indicator): Likewise.
29054         (dr_known_forward_stride_p): Likewise.
29055         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
29056         tree-ssanames.h.
29057         (runtime_alias_check_p): Allow runtime alias checks with
29058         variable strides.
29059         (operator ==): Compare access_size and align.
29060         (prune_runtime_alias_test_list): Rework for new distinction between
29061         the access_size and seg_len.
29062         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
29063         segment lengths.
29064         (get_segment_min_max): New function.
29065         (create_intersect_range_checks): Use it.
29066         (dr_step_indicator): New function.
29067         (dr_direction_indicator): Likewise.
29068         (dr_zero_step_indicator): Likewise.
29069         (dr_known_forward_stride_p): Likewise.
29070         * tree-loop-distribution.c (data_ref_segment_size): Return
29071         DR_STEP * (niters - 1).
29072         (compute_alias_check_pairs): Update call to the dr_with_seg_len
29073         constructor.
29074         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
29075         (vect_preserves_scalar_order_p): New function, split out from...
29076         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
29077         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
29078         (vect_vfa_access_size): New function.
29079         (vect_vfa_align): Likewise.
29080         (vect_compile_time_alias): Take access_size_a and access_b arguments.
29081         (dump_lower_bound): New function.
29082         (vect_check_lower_bound): Likewise.
29083         (vect_small_gap_p): Likewise.
29084         (vectorizable_with_step_bound_p): Likewise.
29085         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
29086         depencies if the vectorization factor is 1.  Convert the checks
29087         for nonzero steps into checks on the bounds of DR_STEP.  Try using
29088         a bunds check for variable steps if the minimum required step is
29089         relatively small. Update calls to the dr_with_seg_len
29090         constructor and to vect_compile_time_alias.
29091         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
29092         function.
29093         (vect_loop_versioning): Call it.
29094         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
29095         when retrying.
29096         (vect_estimate_min_profitable_iters): Account for any bounds checks.
29098 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29099             Alan Hayward  <alan.hayward@arm.com>
29100             David Sherwood  <david.sherwood@arm.com>
29102         * doc/sourcebuild.texi (vect_scatter_store): Document.
29103         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
29104         optabs.
29105         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
29106         Document.
29107         * genopinit.c (main): Add supports_vec_scatter_store and
29108         supports_vec_scatter_store_cached to target_optabs.
29109         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
29110         IFN_MASK_SCATTER_STORE.
29111         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
29112         functions.
29113         * internal-fn.h (internal_store_fn_p): Declare.
29114         (internal_fn_stored_value_index): Likewise.
29115         * internal-fn.c (scatter_store_direct): New macro.
29116         (expand_scatter_store_optab_fn): New function.
29117         (direct_scatter_store_optab_supported_p): New macro.
29118         (internal_store_fn_p): New function.
29119         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
29120         IFN_MASK_SCATTER_STORE.
29121         (internal_fn_mask_index): Likewise.
29122         (internal_fn_stored_value_index): New function.
29123         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
29124         for scatter stores.
29125         * optabs-query.h (supports_vec_scatter_store_p): Declare.
29126         * optabs-query.c (supports_vec_scatter_store_p): New function.
29127         * tree-vectorizer.h (vect_get_store_rhs): Declare.
29128         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
29129         true for scatter stores.
29130         (vect_gather_scatter_fn_p): Handle scatter stores too.
29131         (vect_check_gather_scatter): Consider using scatter stores if
29132         supports_vec_scatter_store_p.
29133         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
29134         scatter stores too.
29135         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
29136         internal_fn_stored_value_index.
29137         (check_load_store_masking): Handle scatter stores too.
29138         (vect_get_store_rhs): Make public.
29139         (vectorizable_call): Use internal_store_fn_p.
29140         (vectorizable_store): Handle scatter store internal functions.
29141         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
29142         when deciding whether the end of the group has been reached.
29143         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
29144         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
29145         (mask_scatter_store<mode>): New insns.
29147 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29148             Alan Hayward  <alan.hayward@arm.com>
29149             David Sherwood  <david.sherwood@arm.com>
29151         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
29152         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
29153         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
29154         function.
29155         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
29156         Use vect_truncate_gather_scatter_offset if we can't treat the
29157         operation as a normal gather load or scatter store.
29158         (get_group_load_store_type): Take the gather_scatter_info
29159         as argument.  Try using a gather load or scatter store for
29160         single-element groups.
29161         (get_load_store_type): Update calls to get_group_load_store_type
29162         and vect_use_strided_gather_scatters_p.
29164 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29165             Alan Hayward  <alan.hayward@arm.com>
29166             David Sherwood  <david.sherwood@arm.com>
29168         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
29169         optional tree argument.
29170         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
29171         null target hooks.
29172         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
29173         but continue to use the current value as a fallback.
29174         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
29175         to compare the updates.
29176         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
29177         (get_load_store_type): Use it when handling a strided access.
29178         (vect_get_strided_load_store_ops): New function.
29179         (vect_get_data_ptr_increment): Likewise.
29180         (vectorizable_load): Handle strided gather loads.  Always pass
29181         a step to vect_create_data_ref_ptr and bump_vector_ptr.
29183 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29184             Alan Hayward  <alan.hayward@arm.com>
29185             David Sherwood  <david.sherwood@arm.com>
29187         * doc/md.texi (gather_load@var{m}): Document.
29188         (mask_gather_load@var{m}): Likewise.
29189         * genopinit.c (main): Add supports_vec_gather_load and
29190         supports_vec_gather_load_cached to target_optabs.
29191         * optabs-tree.c (init_tree_optimization_optabs): Use
29192         ggc_cleared_alloc to allocate target_optabs.
29193         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
29194         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
29195         functions.
29196         * internal-fn.h (internal_load_fn_p): Declare.
29197         (internal_gather_scatter_fn_p): Likewise.
29198         (internal_fn_mask_index): Likewise.
29199         (internal_gather_scatter_fn_supported_p): Likewise.
29200         * internal-fn.c (gather_load_direct): New macro.
29201         (expand_gather_load_optab_fn): New function.
29202         (direct_gather_load_optab_supported_p): New macro.
29203         (direct_internal_fn_optab): New function.
29204         (internal_load_fn_p): Likewise.
29205         (internal_gather_scatter_fn_p): Likewise.
29206         (internal_fn_mask_index): Likewise.
29207         (internal_gather_scatter_fn_supported_p): Likewise.
29208         * optabs-query.c (supports_at_least_one_mode_p): New function.
29209         (supports_vec_gather_load_p): Likewise.
29210         * optabs-query.h (supports_vec_gather_load_p): Declare.
29211         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
29212         and memory_type field.
29213         (NUM_PATTERNS): Bump to 15.
29214         * tree-vect-data-refs.c: Include internal-fn.h.
29215         (vect_gather_scatter_fn_p): New function.
29216         (vect_describe_gather_scatter_call): Likewise.
29217         (vect_check_gather_scatter): Try using internal functions for
29218         gather loads.  Recognize existing calls to a gather load function.
29219         (vect_analyze_data_refs): Consider using gather loads if
29220         supports_vec_gather_load_p.
29221         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
29222         (vect_get_gather_scatter_offset_type): Likewise.
29223         (vect_convert_mask_for_vectype): Likewise.
29224         (vect_add_conversion_to_patterm): Likewise.
29225         (vect_try_gather_scatter_pattern): Likewise.
29226         (vect_recog_gather_scatter_pattern): New pattern recognizer.
29227         (vect_vect_recog_func_ptrs): Add it.
29228         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
29229         internal_fn_mask_index and internal_gather_scatter_fn_p.
29230         (check_load_store_masking): Take the gather_scatter_info as an
29231         argument and handle gather loads.
29232         (vect_get_gather_scatter_ops): New function.
29233         (vectorizable_call): Check internal_load_fn_p.
29234         (vectorizable_load): Likewise.  Handle gather load internal
29235         functions.
29236         (vectorizable_store): Update call to check_load_store_masking.
29237         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
29238         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
29239         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
29240         (aarch64_gather_scale_operand_d): New predicates.
29241         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
29242         (mask_gather_load<mode>): New insns.
29244 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29245             Alan Hayward  <alan.hayward@arm.com>
29246             David Sherwood  <david.sherwood@arm.com>
29248         * optabs.def (fold_left_plus_optab): New optab.
29249         * doc/md.texi (fold_left_plus_@var{m}): Document.
29250         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
29251         * internal-fn.c (fold_left_direct): Define.
29252         (expand_fold_left_optab_fn): Likewise.
29253         (direct_fold_left_optab_supported_p): Likewise.
29254         * fold-const-call.c (fold_const_fold_left): New function.
29255         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
29256         * tree-parloops.c (valid_reduction_p): New function.
29257         (gather_scalar_reductions): Use it.
29258         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
29259         (vect_finish_replace_stmt): Declare.
29260         * tree-vect-loop.c (fold_left_reduction_fn): New function.
29261         (needs_fold_left_reduction_p): New function, split out from...
29262         (vect_is_simple_reduction): ...here.  Accept reductions that
29263         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
29264         (vect_force_simple_reduction): Also store the reduction type in
29265         the assignment's STMT_VINFO_REDUC_TYPE.
29266         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
29267         (merge_with_identity): New function.
29268         (vect_expand_fold_left): Likewise.
29269         (vectorize_fold_left_reduction): Likewise.
29270         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
29271         scalar phi in place for it.  Check for target support and reject
29272         cases that would reassociate the operation.  Defer the transform
29273         phase to vectorize_fold_left_reduction.
29274         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
29275         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
29276         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
29278 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29280         * tree-if-conv.c (predicate_mem_writes): Remove redundant
29281         call to ifc_temp_var.
29283 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29284             Alan Hayward  <alan.hayward@arm.com>
29285             David Sherwood  <david.sherwood@arm.com>
29287         * target.def (legitimize_address_displacement): Take the original
29288         offset as a poly_int.
29289         * targhooks.h (default_legitimize_address_displacement): Update
29290         accordingly.
29291         * targhooks.c (default_legitimize_address_displacement): Likewise.
29292         * doc/tm.texi: Regenerate.
29293         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
29294         as an argument, moving assert of ad->disp == ad->disp_term to...
29295         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
29296         Try calling targetm.legitimize_address_displacement before expanding
29297         the address rather than afterwards, and adjust for the new interface.
29298         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
29299         Match the new hook interface.  Handle SVE addresses.
29300         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
29301         new hook interface.
29303 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29305         * Makefile.in (OBJS): Add early-remat.o.
29306         * target.def (select_early_remat_modes): New hook.
29307         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
29308         * doc/tm.texi: Regenerate.
29309         * targhooks.h (default_select_early_remat_modes): Declare.
29310         * targhooks.c (default_select_early_remat_modes): New function.
29311         * timevar.def (TV_EARLY_REMAT): New timevar.
29312         * passes.def (pass_early_remat): New pass.
29313         * tree-pass.h (make_pass_early_remat): Declare.
29314         * early-remat.c: New file.
29315         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
29316         function.
29317         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
29319 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29320             Alan Hayward  <alan.hayward@arm.com>
29321             David Sherwood  <david.sherwood@arm.com>
29323         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
29324         vfm1 with a bound_epilog parameter.
29325         (vect_do_peeling): Update calls accordingly, and move the prologue
29326         call earlier in the function.  Treat the base bound_epilog as 0 for
29327         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
29328         this base when peeling for gaps.
29329         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
29330         with fully-masked loops.
29331         (vect_estimate_min_profitable_iters): Handle the single peeled
29332         iteration in that case.
29334 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29335             Alan Hayward  <alan.hayward@arm.com>
29336             David Sherwood  <david.sherwood@arm.com>
29338         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
29339         single-element interleaving even if the size is not a power of 2.
29340         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
29341         accesses for single-element interleaving if the group size is
29342         not a power of 2.
29344 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29345             Alan Hayward  <alan.hayward@arm.com>
29346             David Sherwood  <david.sherwood@arm.com>
29348         * doc/md.texi (fold_extract_last_@var{m}): Document.
29349         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
29350         * optabs.def (fold_extract_last_optab): New optab.
29351         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
29352         * internal-fn.c (fold_extract_direct): New macro.
29353         (expand_fold_extract_optab_fn): Likewise.
29354         (direct_fold_extract_optab_supported_p): Likewise.
29355         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
29356         * tree-vect-loop.c (vect_model_reduction_cost): Handle
29357         EXTRACT_LAST_REDUCTION.
29358         (get_initial_def_for_reduction): Do not create an initial vector
29359         for EXTRACT_LAST_REDUCTION reductions.
29360         (vectorizable_reduction): Leave the scalar phi in place for
29361         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
29362         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
29363         epilogue code for EXTRACT_LAST_REDUCTION and defer the
29364         transform phase to vectorizable_condition.
29365         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
29366         split out from...
29367         (vect_finish_stmt_generation): ...here.
29368         (vect_finish_replace_stmt): New function.
29369         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
29370         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
29371         pattern.
29372         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
29374 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29375             Alan Hayward  <alan.hayward@arm.com>
29376             David Sherwood  <david.sherwood@arm.com>
29378         * doc/md.texi (extract_last_@var{m}): Document.
29379         * optabs.def (extract_last_optab): New optab.
29380         * internal-fn.def (EXTRACT_LAST): New internal function.
29381         * internal-fn.c (cond_unary_direct): New macro.
29382         (expand_cond_unary_optab_fn): Likewise.
29383         (direct_cond_unary_optab_supported_p): Likewise.
29384         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
29385         loops using EXTRACT_LAST.
29386         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
29387         (extract_last_<mode>): ...this optab.
29388         (vec_extract<mode><Vel>): Update accordingly.
29390 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29391             Alan Hayward  <alan.hayward@arm.com>
29392             David Sherwood  <david.sherwood@arm.com>
29394         * target.def (empty_mask_is_expensive): New hook.
29395         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
29396         * doc/tm.texi: Regenerate.
29397         * targhooks.h (default_empty_mask_is_expensive): Declare.
29398         * targhooks.c (default_empty_mask_is_expensive): New function.
29399         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
29400         if the target says that empty masks are expensive.
29401         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
29402         New function.
29403         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
29405 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29406             Alan Hayward  <alan.hayward@arm.com>
29407             David Sherwood  <david.sherwood@arm.com>
29409         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
29410         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
29411         (vect_use_loop_mask_for_alignment_p): New function.
29412         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
29413         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
29414         niters_skip argument.  Make sure that the first niters_skip elements
29415         of the first iteration are inactive.
29416         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
29417         Update call to vect_set_loop_masks_directly.
29418         (get_misalign_in_elems): New function, split out from...
29419         (vect_gen_prolog_loop_niters): ...here.
29420         (vect_update_init_of_dr): Take a code argument that specifies whether
29421         the adjustment should be added or subtracted.
29422         (vect_update_init_of_drs): Likewise.
29423         (vect_prepare_for_masked_peels): New function.
29424         (vect_do_peeling): Skip prologue peeling if we're using a mask
29425         instead.  Update call to vect_update_inits_of_drs.
29426         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29427         mask_skip_niters.
29428         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
29429         alignment.  Do not include the number of peeled iterations in
29430         the minimum threshold in that case.
29431         (vectorizable_induction): Adjust the start value down by
29432         LOOP_VINFO_MASK_SKIP_NITERS iterations.
29433         (vect_transform_loop): Call vect_prepare_for_masked_peels.
29434         Take the number of skipped iterations into account when calculating
29435         the loop bounds.
29436         * tree-vect-stmts.c (vect_gen_while_not): New function.
29438 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29439             Alan Hayward  <alan.hayward@arm.com>
29440             David Sherwood  <david.sherwood@arm.com>
29442         * doc/sourcebuild.texi (vect_fully_masked): Document.
29443         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
29444         default value to 0.
29445         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
29446         split out from...
29447         (vect_analyze_loop_2): ...here. Don't check the vectorization
29448         factor against the number of loop iterations if the loop is
29449         fully-masked.
29451 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29452             Alan Hayward  <alan.hayward@arm.com>
29453             David Sherwood  <david.sherwood@arm.com>
29455         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
29456         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
29457         (dump_groups): Update accordingly.
29458         (iv_use::mem_type): New member variable.
29459         (address_p): New function.
29460         (record_use): Add a mem_type argument and initialize the new
29461         mem_type field.
29462         (record_group_use): Add a mem_type argument.  Use address_p.
29463         Remove obsolete null checks of base_object.  Update call to record_use.
29464         (find_interesting_uses_op): Update call to record_group_use.
29465         (find_interesting_uses_cond): Likewise.
29466         (find_interesting_uses_address): Likewise.
29467         (get_mem_type_for_internal_fn): New function.
29468         (find_address_like_use): Likewise.
29469         (find_interesting_uses_stmt): Try find_address_like_use before
29470         calling find_interesting_uses_op.
29471         (addr_offset_valid_p): Use the iv mem_type field as the type
29472         of the addressed memory.
29473         (add_autoinc_candidates): Likewise.
29474         (get_address_cost): Likewise.
29475         (split_small_address_groups_p): Use address_p.
29476         (split_address_groups): Likewise.
29477         (add_iv_candidate_for_use): Likewise.
29478         (autoinc_possible_for_pair): Likewise.
29479         (rewrite_groups): Likewise.
29480         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
29481         (determine_group_iv_cost): Update after split of USE_ADDRESS.
29482         (get_alias_ptr_type_for_ptr_address): New function.
29483         (rewrite_use_address): Rewrite address uses in calls that were
29484         identified by find_address_like_use.
29486 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29487             Alan Hayward  <alan.hayward@arm.com>
29488             David Sherwood  <david.sherwood@arm.com>
29490         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
29491         TARGET_MEM_REFs.
29492         * gimple-expr.h (is_gimple_addressable: Likewise.
29493         * gimple-expr.c (is_gimple_address): Likewise.
29494         * internal-fn.c (expand_call_mem_ref): New function.
29495         (expand_mask_load_optab_fn): Use it.
29496         (expand_mask_store_optab_fn): Likewise.
29498 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29499             Alan Hayward  <alan.hayward@arm.com>
29500             David Sherwood  <david.sherwood@arm.com>
29502         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
29503         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
29504         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
29505         (cond_umax@var{mode}): Document.
29506         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
29507         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
29508         (cond_umin_optab, cond_umax_optab): New optabs.
29509         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
29510         (COND_IOR, COND_XOR): New internal functions.
29511         * internal-fn.h (get_conditional_internal_fn): Declare.
29512         * internal-fn.c (cond_binary_direct): New macro.
29513         (expand_cond_binary_optab_fn): Likewise.
29514         (direct_cond_binary_optab_supported_p): Likewise.
29515         (get_conditional_internal_fn): New function.
29516         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
29517         Cope with reduction statements that are vectorized as calls rather
29518         than assignments.
29519         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
29520         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
29521         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
29522         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
29523         (UNSPEC_COND_EOR): New unspecs.
29524         (optab): Add mappings for them.
29525         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
29526         (sve_int_op, sve_fp_op): New int attributes.
29528 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29529             Alan Hayward  <alan.hayward@arm.com>
29530             David Sherwood  <david.sherwood@arm.com>
29532         * optabs.def (while_ult_optab): New optab.
29533         * doc/md.texi (while_ult@var{m}@var{n}): Document.
29534         * internal-fn.def (WHILE_ULT): New internal function.
29535         * internal-fn.h (direct_internal_fn_supported_p): New override
29536         that takes two types as argument.
29537         * internal-fn.c (while_direct): New macro.
29538         (expand_while_optab_fn): New function.
29539         (convert_optab_supported_p): Likewise.
29540         (direct_while_optab_supported_p): New macro.
29541         * wide-int.h (wi::udiv_ceil): New function.
29542         * tree-vectorizer.h (rgroup_masks): New structure.
29543         (vec_loop_masks): New typedef.
29544         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
29545         and fully_masked_p.
29546         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
29547         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
29548         (vect_max_vf): New function.
29549         (slpeel_make_loop_iterate_ntimes): Delete.
29550         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
29551         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
29552         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
29553         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
29554         internal-fn.h, stor-layout.h and optabs-query.h.
29555         (vect_set_loop_mask): New function.
29556         (add_preheader_seq): Likewise.
29557         (add_header_seq): Likewise.
29558         (interleave_supported_p): Likewise.
29559         (vect_maybe_permute_loop_masks): Likewise.
29560         (vect_set_loop_masks_directly): Likewise.
29561         (vect_set_loop_condition_masked): Likewise.
29562         (vect_set_loop_condition_unmasked): New function, split out from
29563         slpeel_make_loop_iterate_ntimes.
29564         (slpeel_make_loop_iterate_ntimes): Rename to..
29565         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
29566         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
29567         (vect_do_peeling): Update call accordingly.
29568         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
29569         loops.
29570         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29571         mask_compare_type, can_fully_mask_p and fully_masked_p.
29572         (release_vec_loop_masks): New function.
29573         (_loop_vec_info): Use it to free the loop masks.
29574         (can_produce_all_loop_masks_p): New function.
29575         (vect_get_max_nscalars_per_iter): Likewise.
29576         (vect_verify_full_masking): Likewise.
29577         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
29578         retries, and free the mask rgroups before retrying.  Check loop-wide
29579         reasons for disallowing fully-masked loops.  Make the final decision
29580         about whether use a fully-masked loop or not.
29581         (vect_estimate_min_profitable_iters): Do not assume that peeling
29582         for the number of iterations will be needed for fully-masked loops.
29583         (vectorizable_reduction): Disable fully-masked loops.
29584         (vectorizable_live_operation): Likewise.
29585         (vect_halve_mask_nunits): New function.
29586         (vect_double_mask_nunits): Likewise.
29587         (vect_record_loop_mask): Likewise.
29588         (vect_get_loop_mask): Likewise.
29589         (vect_transform_loop): Handle the case in which the final loop
29590         iteration might handle a partial vector.  Call vect_set_loop_condition
29591         instead of slpeel_make_loop_iterate_ntimes.
29592         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
29593         (check_load_store_masking): New function.
29594         (prepare_load_store_mask): Likewise.
29595         (vectorizable_store): Handle fully-masked loops.
29596         (vectorizable_load): Likewise.
29597         (supportable_widening_operation): Use vect_halve_mask_nunits for
29598         booleans.
29599         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
29600         (vect_gen_while): New function.
29601         * config/aarch64/aarch64.md (umax<mode>3): New expander.
29602         (aarch64_uqdec<mode>): New insn.
29604 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29605             Alan Hayward  <alan.hayward@arm.com>
29606             David Sherwood  <david.sherwood@arm.com>
29608         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
29609         (reduc_xor_scal_optab): New optabs.
29610         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
29611         (reduc_xor_scal_@var{m}): Document.
29612         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
29613         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
29614         internal functions.
29615         * fold-const-call.c (fold_const_call): Handle them.
29616         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
29617         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
29618         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
29619         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
29620         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
29621         (UNSPEC_XORV): New unspecs.
29622         (optab): Add entries for them.
29623         (BITWISEV): New int iterator.
29624         (bit_reduc_op): New int attributes.
29626 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29627             Alan Hayward  <alan.hayward@arm.com>
29628             David Sherwood  <david.sherwood@arm.com>
29630         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
29631         * internal-fn.def (VEC_SHL_INSERT): New internal function.
29632         * optabs.def (vec_shl_insert_optab): New optab.
29633         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
29634         (duplicate_and_interleave): Likewise.
29635         * tree-vect-loop.c: Include internal-fn.h.
29636         (neutral_op_for_slp_reduction): New function, split out from
29637         get_initial_defs_for_reduction.
29638         (get_initial_def_for_reduction): Handle option 2 for variable-length
29639         vectors by loading the neutral value into a vector and then shifting
29640         the initial value into element 0.
29641         (get_initial_defs_for_reduction): Replace the code argument with
29642         the neutral value calculated by neutral_op_for_slp_reduction.
29643         Use gimple_build_vector for constant-length vectors.
29644         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
29645         but the first group_size elements have a neutral value.
29646         Use duplicate_and_interleave otherwise.
29647         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
29648         Update call to get_initial_defs_for_reduction.  Handle SLP
29649         reductions for variable-length vectors by creating one vector
29650         result for each scalar result, with the elements associated
29651         with other scalar results stubbed out with the neutral value.
29652         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
29653         Require IFN_VEC_SHL_INSERT for double reductions on
29654         variable-length vectors, or SLP reductions that have
29655         a neutral value.  Require can_duplicate_and_interleave_p
29656         support for variable-length unchained SLP reductions if there
29657         is no neutral value, such as for MIN/MAX reductions.  Also require
29658         the number of vector elements to be a multiple of the number of
29659         SLP statements when doing variable-length unchained SLP reductions.
29660         Update call to vect_create_epilog_for_reduction.
29661         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
29662         and remove initial values.
29663         (duplicate_and_interleave): Make public.
29664         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
29665         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
29667 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29668             Alan Hayward  <alan.hayward@arm.com>
29669             David Sherwood  <david.sherwood@arm.com>
29671         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
29672         (can_duplicate_and_interleave_p): New function.
29673         (vect_get_and_check_slp_defs): Take the vector of statements
29674         rather than just the current one.  Remove excess parentheses.
29675         Restriction rejectinon of vect_constant_def and vect_external_def
29676         for variable-length vectors to boolean types, or types for which
29677         can_duplicate_and_interleave_p is false.
29678         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
29679         (duplicate_and_interleave): New function.
29680         (vect_get_constant_vectors): Use gimple_build_vector for
29681         constant-length vectors and suitable variable-length constant
29682         vectors.  Use duplicate_and_interleave for other variable-length
29683         vectors.  Don't defer the update when inserting new statements.
29685 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29686             Alan Hayward  <alan.hayward@arm.com>
29687             David Sherwood  <david.sherwood@arm.com>
29689         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29690         min_profitable_iters doesn't go negative.
29692 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29693             Alan Hayward  <alan.hayward@arm.com>
29694             David Sherwood  <david.sherwood@arm.com>
29696         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29697         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29698         * optabs.def (vec_mask_load_lanes_optab): New optab.
29699         (vec_mask_store_lanes_optab): Likewise.
29700         * internal-fn.def (MASK_LOAD_LANES): New internal function.
29701         (MASK_STORE_LANES): Likewise.
29702         * internal-fn.c (mask_load_lanes_direct): New macro.
29703         (mask_store_lanes_direct): Likewise.
29704         (expand_mask_load_optab_fn): Handle masked operations.
29705         (expand_mask_load_lanes_optab_fn): New macro.
29706         (expand_mask_store_optab_fn): Handle masked operations.
29707         (expand_mask_store_lanes_optab_fn): New macro.
29708         (direct_mask_load_lanes_optab_supported_p): Likewise.
29709         (direct_mask_store_lanes_optab_supported_p): Likewise.
29710         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29711         parameter.
29712         (vect_load_lanes_supported): Likewise.
29713         * tree-vect-data-refs.c (strip_conversion): New function.
29714         (can_group_stmts_p): Likewise.
29715         (vect_analyze_data_ref_accesses): Use it instead of checking
29716         for a pair of assignments.
29717         (vect_store_lanes_supported): Take a masked_p parameter.
29718         (vect_load_lanes_supported): Likewise.
29719         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29720         vect_store_lanes_supported and vect_load_lanes_supported.
29721         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29722         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29723         parameter.  Don't allow gaps for masked accesses.
29724         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
29725         and vect_load_lanes_supported.
29726         (get_load_store_type): Take a masked_p parameter and update
29727         call to get_group_load_store_type.
29728         (vectorizable_store): Update call to get_load_store_type.
29729         Handle IFN_MASK_STORE_LANES.
29730         (vectorizable_load): Update call to get_load_store_type.
29731         Handle IFN_MASK_LOAD_LANES.
29733 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29734             Alan Hayward  <alan.hayward@arm.com>
29735             David Sherwood  <david.sherwood@arm.com>
29737         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29738         modes for SVE.
29739         * config/aarch64/aarch64-protos.h
29740         (aarch64_sve_struct_memory_operand_p): Declare.
29741         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29742         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29743         (VPRED, vpred): Handle SVE structure modes.
29744         * config/aarch64/constraints.md (Utx): New constraint.
29745         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29746         (aarch64_sve_struct_nonimmediate_operand): New predicates.
29747         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29748         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29749         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29750         structure modes.  Split into pieces after RA.
29751         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29752         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29753         New patterns.
29754         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29755         SVE structure modes.
29756         (aarch64_classify_address): Likewise.
29757         (sizetochar): Move earlier in file.
29758         (aarch64_print_operand): Handle SVE register lists.
29759         (aarch64_array_mode): New function.
29760         (aarch64_sve_struct_memory_operand_p): Likewise.
29761         (TARGET_ARRAY_MODE): Redefine.
29763 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29764             Alan Hayward  <alan.hayward@arm.com>
29765             David Sherwood  <david.sherwood@arm.com>
29767         * target.def (array_mode): New target hook.
29768         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29769         * doc/tm.texi: Regenerate.
29770         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29771         * hooks.c (hook_optmode_mode_uhwi_none): New function.
29772         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29773         targetm.array_mode.
29774         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
29775         type sizes.
29777 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29778             Alan Hayward  <alan.hayward@arm.com>
29779             David Sherwood  <david.sherwood@arm.com>
29781         * fold-const.c (fold_binary_loc): Check the argument types
29782         rather than the result type when testing for a vector operation.
29784 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29786         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29787         * doc/tm.texi: Regenerate.
29789 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29790             Alan Hayward  <alan.hayward@arm.com>
29791             David Sherwood  <david.sherwood@arm.com>
29793         * doc/invoke.texi (-msve-vector-bits=): Document new option.
29794         (sve): Document new AArch64 extension.
29795         * doc/md.texi (w): Extend the description of the AArch64
29796         constraint to include SVE vectors.
29797         (Upl, Upa): Document new AArch64 predicate constraints.
29798         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29799         enum.
29800         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29801         (msve-vector-bits=): New option.
29802         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29803         SVE when these are disabled.
29804         (sve): New extension.
29805         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29806         modes.  Adjust their number of units based on aarch64_sve_vg.
29807         (MAX_BITSIZE_MODE_ANY_MODE): Define.
29808         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29809         aarch64_addr_query_type.
29810         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29811         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29812         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29813         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29814         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29815         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29816         (aarch64_simd_imm_zero_p): Delete.
29817         (aarch64_check_zero_based_sve_index_immediate): Declare.
29818         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29819         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29820         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29821         (aarch64_sve_float_mul_immediate_p): Likewise.
29822         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29823         rather than an rtx.
29824         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29825         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29826         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29827         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29828         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29829         (aarch64_regmode_natural_size): Likewise.
29830         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29831         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29832         left one place.
29833         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29834         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29835         for VG and the SVE predicate registers.
29836         (V_ALIASES): Add a "z"-prefixed alias.
29837         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29838         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29839         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29840         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29841         (REG_CLASS_NAMES): Add entries for them.
29842         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29843         and the predicate registers.
29844         (aarch64_sve_vg): Declare.
29845         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29846         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29847         (REGMODE_NATURAL_SIZE): Define.
29848         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29849         SVE macros.
29850         * config/aarch64/aarch64.c: Include cfgrtl.h.
29851         (simd_immediate_info): Add a constructor for series vectors,
29852         and an associated step field.
29853         (aarch64_sve_vg): New variable.
29854         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29855         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29856         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29857         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29858         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29859         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29860         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29861         (aarch64_get_mask_mode): New functions.
29862         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29863         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29864         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29865         predicate modes and predicate registers.  Explicitly restrict
29866         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29867         to store a vector mode if it is recognized by
29868         aarch64_classify_vector_mode.
29869         (aarch64_regmode_natural_size): New function.
29870         (aarch64_hard_regno_caller_save_mode): Return the original mode
29871         for predicates.
29872         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29873         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29874         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29875         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29876         functions.
29877         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29878         does not overlap dest if the function is frame-related.  Handle
29879         SVE constants.
29880         (aarch64_split_add_offset): New function.
29881         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29882         them aarch64_add_offset.
29883         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29884         and update call to aarch64_sub_sp.
29885         (aarch64_add_cfa_expression): New function.
29886         (aarch64_expand_prologue): Pass extra temporary registers to the
29887         functions above.  Handle the case in which we need to emit new
29888         DW_CFA_expressions for registers that were originally saved
29889         relative to the stack pointer, but now have to be expressed
29890         relative to the frame pointer.
29891         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29892         functions above.
29893         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29894         IP0 and IP1 values for SVE frames.
29895         (aarch64_expand_vec_series): New function.
29896         (aarch64_expand_sve_widened_duplicate): Likewise.
29897         (aarch64_expand_sve_const_vector): Likewise.
29898         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29899         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29900         into the register, rather than emitting a SET directly.
29901         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29902         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29903         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29904         (offset_9bit_signed_scaled_p): New functions.
29905         (aarch64_replicate_bitmask_imm): New function.
29906         (aarch64_bitmask_imm): Use it.
29907         (aarch64_cannot_force_const_mem): Reject expressions involving
29908         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29909         (aarch64_classify_index): Handle SVE indices, by requiring
29910         a plain register index with a scale that matches the element size.
29911         (aarch64_classify_address): Handle SVE addresses.  Assert that
29912         the mode of the address is VOIDmode or an integer mode.
29913         Update call to aarch64_classify_symbol.
29914         (aarch64_classify_symbolic_expression): Update call to
29915         aarch64_classify_symbol.
29916         (aarch64_const_vec_all_in_range_p): New function.
29917         (aarch64_print_vector_float_operand): Likewise.
29918         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29919         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29920         and the FP immediates 1.0 and 0.5.
29921         (aarch64_print_address_internal): Handle SVE addresses.
29922         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29923         (aarch64_regno_regclass): Handle predicate registers.
29924         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29925         data modes.
29926         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29927         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29928         (aarch64_convert_sve_vector_bits): New function.
29929         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29930         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29931         rather than an rtx.
29932         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29933         Handle SVE vector and predicate modes.  Accept VL-based constants
29934         that need only one temporary register, and VL offsets that require
29935         no temporary registers.
29936         (aarch64_conditional_register_usage): Mark the predicate registers
29937         as fixed if SVE isn't available.
29938         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29939         Return true for SVE vector and predicate modes.
29940         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29941         rather than an unsigned int.  Handle SVE modes.
29942         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29943         SVE modes.
29944         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29945         if SVE is enabled.
29946         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29947         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29948         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29949         (aarch64_sve_float_mul_immediate_p): New functions.
29950         (aarch64_sve_valid_immediate): New function.
29951         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29952         Explicitly reject structure modes.  Check for INDEX constants.
29953         Handle PTRUE and PFALSE constants.
29954         (aarch64_check_zero_based_sve_index_immediate): New function.
29955         (aarch64_simd_imm_zero_p): Delete.
29956         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29957         vector modes.  Accept constants in the range of CNT[BHWD].
29958         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29959         ask for an Advanced SIMD mode.
29960         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29961         (aarch64_simd_vector_alignment): Handle SVE predicates.
29962         (aarch64_vectorize_preferred_vector_alignment): New function.
29963         (aarch64_simd_vector_alignment_reachable): Use it instead of
29964         the vector size.
29965         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29966         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29967         functions.
29968         (MAX_VECT_LEN): Delete.
29969         (expand_vec_perm_d): Add a vec_flags field.
29970         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29971         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29972         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29973         for SVE modes.
29974         (aarch64_evpc_rev): Rename to...
29975         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29976         (aarch64_evpc_rev_global): New function.
29977         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29978         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29979         MAX_VECT_LEN.
29980         (aarch64_evpc_sve_tbl): New function.
29981         (aarch64_expand_vec_perm_const_1): Update after rename of
29982         aarch64_evpc_rev.  Handle SVE permutes too, trying
29983         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29984         than aarch64_evpc_tbl.
29985         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29986         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29987         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29988         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29989         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29990         (aarch64_expand_sve_vcond): New functions.
29991         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29992         of aarch64_vector_mode_p.
29993         (aarch64_dwarf_poly_indeterminate_value): New function.
29994         (aarch64_compute_pressure_classes): Likewise.
29995         (aarch64_can_change_mode_class): Likewise.
29996         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29997         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29998         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29999         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
30000         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
30001         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
30002         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
30003         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
30004         constraints.
30005         (Dn, Dl, Dr): Accept const as well as const_vector.
30006         (Dz): Likewise.  Compare against CONST0_RTX.
30007         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
30008         of "vector" where appropriate.
30009         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
30010         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
30011         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
30012         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
30013         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
30014         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
30015         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
30016         (v_int_equiv): Extend to SVE modes.
30017         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
30018         mode attributes.
30019         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
30020         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
30021         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
30022         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
30023         (SVE_COND_FP_CMP): New int iterators.
30024         (perm_hilo): Handle the new unpack unspecs.
30025         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
30026         attributes.
30027         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
30028         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
30029         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
30030         (aarch64_equality_operator, aarch64_constant_vector_operand)
30031         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
30032         (aarch64_sve_nonimmediate_operand): Likewise.
30033         (aarch64_sve_general_operand): Likewise.
30034         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
30035         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
30036         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
30037         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
30038         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
30039         (aarch64_sve_float_arith_immediate): Likewise.
30040         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
30041         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
30042         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
30043         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
30044         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
30045         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
30046         (aarch64_sve_float_arith_operand): Likewise.
30047         (aarch64_sve_float_arith_with_sub_operand): Likewise.
30048         (aarch64_sve_float_mul_operand): Likewise.
30049         (aarch64_sve_vec_perm_operand): Likewise.
30050         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
30051         (aarch64_mov_operand): Accept const_poly_int and const_vector.
30052         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
30053         as well as const_vector.
30054         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
30055         in file.  Use CONST0_RTX and CONSTM1_RTX.
30056         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
30057         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
30058         Use aarch64_simd_imm_zero.
30059         * config/aarch64/aarch64-sve.md: New file.
30060         * config/aarch64/aarch64.md: Include it.
30061         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
30062         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
30063         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
30064         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
30065         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
30066         (sve): New attribute.
30067         (enabled): Disable instructions with the sve attribute unless
30068         TARGET_SVE.
30069         (movqi, movhi): Pass CONST_POLY_INT operaneds through
30070         aarch64_expand_mov_immediate.
30071         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
30072         CNT[BHSD] immediates.
30073         (movti): Split CONST_POLY_INT moves into two halves.
30074         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
30075         Split additions that need a temporary here if the destination
30076         is the stack pointer.
30077         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
30078         (*add<mode>3_poly_1): New instruction.
30079         (set_clobber_cc): New expander.
30081 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30083         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
30084         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
30085         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
30086         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
30087         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
30088         Change innermode from fixed_mode_size to machine_mode.
30089         (simplify_subreg): Update call accordingly.  Handle a constant-sized
30090         subreg of a variable-length CONST_VECTOR.
30092 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30093             Alan Hayward  <alan.hayward@arm.com>
30094             David Sherwood  <david.sherwood@arm.com>
30096         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
30097         (add_offset_to_base): New function, split out from...
30098         (create_mem_ref): ...here.  When handling a scale other than 1,
30099         check first whether the address is valid without the offset.
30100         Add it into the base if so, leaving the index and scale as-is.
30102 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
30104         PR c++/83778
30105         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
30106         fold_for_warn before checking if arg2 is INTEGER_CST.
30108 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
30110         * config/rs6000/predicates.md (load_multiple_operation): Delete.
30111         (store_multiple_operation): Delete.
30112         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
30113         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
30114         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
30115         guarded by TARGET_STRING.
30116         (rs6000_output_load_multiple): Delete.
30117         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30118         OPTION_MASK_STRING / TARGET_STRING handling.
30119         (print_operand) <'N', 'O'>: Add comment that these are unused now.
30120         (const rs6000_opt_masks) <"string">: Change mask to 0.
30121         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
30122         (MASK_STRING): Delete.
30123         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
30124         parts.  Simplify.
30125         (load_multiple): Delete.
30126         (*ldmsi8): Delete.
30127         (*ldmsi7): Delete.
30128         (*ldmsi6): Delete.
30129         (*ldmsi5): Delete.
30130         (*ldmsi4): Delete.
30131         (*ldmsi3): Delete.
30132         (store_multiple): Delete.
30133         (*stmsi8): Delete.
30134         (*stmsi7): Delete.
30135         (*stmsi6): Delete.
30136         (*stmsi5): Delete.
30137         (*stmsi4): Delete.
30138         (*stmsi3): Delete.
30139         (movmemsi_8reg): Delete.
30140         (corresponding unnamed define_insn): Delete.
30141         (movmemsi_6reg): Delete.
30142         (corresponding unnamed define_insn): Delete.
30143         (movmemsi_4reg): Delete.
30144         (corresponding unnamed define_insn): Delete.
30145         (movmemsi_2reg): Delete.
30146         (corresponding unnamed define_insn): Delete.
30147         (movmemsi_1reg): Delete.
30148         (corresponding unnamed define_insn): Delete.
30149         * config/rs6000/rs6000.opt (mno-string): New.
30150         (mstring): Replace by deprecation warning stub.
30151         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
30153 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
30155         * regrename.c (regrename_do_replace): If replacing the same
30156         reg multiple times, try to reuse last created gen_raw_REG.
30158         PR debug/81155
30159         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
30160         main to workaround a bug in GDB.
30162 2018-01-12  Tom de Vries  <tom@codesourcery.com>
30164         PR target/83737
30165         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
30167 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
30169         PR rtl-optimization/80481
30170         * ira-color.c (get_cap_member): New function.
30171         (allocnos_conflict_by_live_ranges_p): Use it.
30172         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
30173         (setup_slot_coalesced_allocno_live_ranges): Ditto.
30175 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
30177         PR target/83628
30178         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
30179         (*saddl_se_1): Ditto.
30180         (*ssubsi_1): Ditto.
30181         (*ssubl_se_1): Ditto.
30183 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30185         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
30186         rather than wi::to_widest for DR_INITs.
30187         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
30188         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
30189         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
30190         INTEGER_CSTs.
30191         (vect_analyze_group_access_1): Note that here.
30193 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30195         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
30196         polynomial type sizes.
30198 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30200         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
30201         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
30202         (gimple_add_tmp_var): Likewise.
30204 2018-01-12  Martin Liska  <mliska@suse.cz>
30206         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
30207         (gimple_alloc_sizes): Likewise.
30208         (dump_gimple_statistics): Use PRIu64 in printf format.
30209         * gimple.h: Change uint64_t to int.
30211 2018-01-12  Martin Liska  <mliska@suse.cz>
30213         * tree-core.h: Use uint64_t instead of int.
30214         * tree.c (tree_node_counts): Likewise.
30215         (tree_node_sizes): Likewise.
30216         (dump_tree_statistics): Use PRIu64 in printf format.
30218 2018-01-12  Martin Liska  <mliska@suse.cz>
30220         * Makefile.in: As qsort_chk is implemented in vec.c, add
30221         vec.o to linkage of gencfn-macros.
30222         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
30223         passing the info to record_node_allocation_statistics.
30224         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
30225         and pass the info.
30226         * ggc-common.c (struct ggc_usage): Add operator== and use
30227         it in operator< and compare function.
30228         * mem-stats.h (struct mem_usage): Likewise.
30229         * vec.c (struct vec_usage): Remove operator< and compare
30230         function. Can be simply inherited.
30232 2018-01-12  Martin Jambor  <mjambor@suse.cz>
30234         PR target/81616
30235         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
30236         * tree-ssa-math-opts.c: Include domwalk.h.
30237         (convert_mult_to_fma_1): New function.
30238         (fma_transformation_info): New type.
30239         (fma_deferring_state): Likewise.
30240         (cancel_fma_deferring): New function.
30241         (result_of_phi): Likewise.
30242         (last_fma_candidate_feeds_initial_phi): Likewise.
30243         (convert_mult_to_fma): Added deferring logic, split actual
30244         transformation to convert_mult_to_fma_1.
30245         (math_opts_dom_walker): New type.
30246         (math_opts_dom_walker::after_dom_children): New method, body moved
30247         here from pass_optimize_widening_mul::execute, added deferring logic
30248         bits.
30249         (pass_optimize_widening_mul::execute): Moved most of code to
30250         math_opts_dom_walker::after_dom_children.
30251         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
30252         * config/i386/i386.c (ix86_option_override_internal): Added
30253         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
30255 2018-01-12  Richard Biener  <rguenther@suse.de>
30257         PR debug/83157
30258         * dwarf2out.c (gen_variable_die): Do not reset old_die for
30259         inline instance vars.
30261 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
30263         PR target/81819
30264         * config/rx/rx.c (rx_is_restricted_memory_address):
30265         Handle SUBREG case.
30267 2018-01-12  Richard Biener  <rguenther@suse.de>
30269         PR tree-optimization/80846
30270         * target.def (split_reduction): New target hook.
30271         * targhooks.c (default_split_reduction): New function.
30272         * targhooks.h (default_split_reduction): Declare.
30273         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
30274         target requests first reduce vectors by combining low and high
30275         parts.
30276         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
30277         (get_vectype_for_scalar_type_and_size): Export.
30278         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
30279         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
30280         * doc/tm.texi: Regenerate.
30281         * config/i386/i386.c (ix86_split_reduction): Implement
30282         TARGET_VECTORIZE_SPLIT_REDUCTION.
30284 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
30286         PR target/83368
30287         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
30288         in PIC mode except for TARGET_VXWORKS_RTP.
30289         * config/sparc/sparc.c: Include cfgrtl.h.
30290         (TARGET_INIT_PIC_REG): Define.
30291         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
30292         (sparc_pic_register_p): New predicate.
30293         (sparc_legitimate_address_p): Use it.
30294         (sparc_legitimize_pic_address): Likewise.
30295         (sparc_delegitimize_address): Likewise.
30296         (sparc_mode_dependent_address_p): Likewise.
30297         (gen_load_pcrel_sym): Remove 4th parameter.
30298         (load_got_register): Adjust call to above.  Remove obsolete stuff.
30299         (sparc_expand_prologue): Do not call load_got_register here.
30300         (sparc_flat_expand_prologue): Likewise.
30301         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
30302         (sparc_use_pseudo_pic_reg): New function.
30303         (sparc_init_pic_reg): Likewise.
30304         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
30305         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
30307 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
30309         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
30310         Add item for branch_cost.
30312 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
30314         PR rtl-optimization/83565
30315         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
30316         not extend the result to a larger mode for rotate operations.
30317         (num_sign_bit_copies1): Likewise.
30319 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30321         PR target/40411
30322         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
30323         -symbolic.
30324         Use values-Xc.o for -pedantic.
30325         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
30327 2018-01-12  Martin Liska  <mliska@suse.cz>
30329         PR ipa/83054
30330         * ipa-devirt.c (final_warning_record::grow_type_warnings):
30331         New function.
30332         (possible_polymorphic_call_targets): Use it.
30333         (ipa_devirt): Likewise.
30335 2018-01-12  Martin Liska  <mliska@suse.cz>
30337         * profile-count.h (enum profile_quality): Use 0 as invalid
30338         enum value of profile_quality.
30340 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
30342         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
30343         -mext-string options.
30345 2018-01-12  Richard Biener  <rguenther@suse.de>
30347         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
30348         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
30349         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
30350         Likewise.
30351         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
30353 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
30355         * configure.ac (--with-long-double-format): Add support for the
30356         configuration option to change the default long double format on
30357         PowerPC systems.
30358         * config.gcc (powerpc*-linux*-*): Likewise.
30359         * configure: Regenerate.
30360         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
30361         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
30362         used without modification.
30364 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30366         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
30367         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
30368         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
30369         MISC_BUILTIN_SPEC_BARRIER.
30370         (rs6000_init_builtins): Likewise.
30371         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
30372         enum value.
30373         (speculation_barrier): New define_insn.
30374         * doc/extend.texi: Document __builtin_speculation_barrier.
30376 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
30378         PR target/83203
30379         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
30380         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
30381         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
30382         iterators.
30383         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
30384         integral modes instead of "ss" and "sd".
30385         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
30386         vectors with 32-bit and 64-bit elements.
30387         (vecdupssescalarmodesuffix): New mode attribute.
30388         (vec_dup<mode>): Use it.
30390 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
30392         PR target/83330
30393         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
30394         frame if argument is passed on stack.
30396 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
30398         PR target/82682
30399         * ree.c (combine_reaching_defs): Optimize also
30400         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
30401         reg2=any_extend(exp); reg1=reg2;, formatting fix.
30403 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
30405         PR middle-end/83189
30406         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
30408 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
30410         PR middle-end/83718
30411         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
30412         after they are computed.
30414 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
30416         PR tree-optimization/83695
30417         * gimple-loop-linterchange.cc
30418         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
30419         reset cached scev information after interchange.
30420         (pass_linterchange::execute): Remove call to scev_reset_htab.
30422 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30424         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
30425         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
30426         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
30427         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
30428         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
30429         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
30430         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
30431         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
30432         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
30433         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
30434         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
30435         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
30436         (V_lane_reg): Likewise.
30437         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
30438         New define_expand.
30439         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
30440         (vfmal_lane_low<mode>_intrinsic,
30441         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
30442         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
30443         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
30444         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
30445         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
30446         vfmsl_lane_high<mode>_intrinsic): New define_insns.
30448 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30450         * config/arm/arm-cpus.in (fp16fml): New feature.
30451         (ALL_SIMD): Add fp16fml.
30452         (armv8.2-a): Add fp16fml as an option.
30453         (armv8.3-a): Likewise.
30454         (armv8.4-a): Add fp16fml as part of fp16.
30455         * config/arm/arm.h (TARGET_FP16FML): Define.
30456         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
30457         when appropriate.
30458         * config/arm/arm-modes.def (V2HF): Define.
30459         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
30460         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
30461         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
30462         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
30463         vfmsl_low, vfmsl_high): New set of builtins.
30464         * config/arm/iterators.md (PLUSMINUS): New code iterator.
30465         (vfml_op): New code attribute.
30466         (VFMLHALVES): New int iterator.
30467         (VFML, VFMLSEL): New mode attributes.
30468         (V_reg): Define mapping for V2HF.
30469         (V_hi, V_lo): New mode attributes.
30470         (VF_constraint): Likewise.
30471         (vfml_half, vfml_half_selector): New int attributes.
30472         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
30473         define_expand.
30474         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
30475         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
30476         New define_insn.
30477         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
30478         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
30479         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
30480         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
30481         documentation.
30482         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
30483         Document new effective target and option set.
30485 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30487         * config/arm/arm-cpus.in (armv8_4): New feature.
30488         (ARMv8_4a): New fgroup.
30489         (armv8.4-a): New arch.
30490         * config/arm/arm-tables.opt: Regenerate.
30491         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
30492         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
30493         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
30494         Add matching rules for -march=armv8.4-a and extensions.
30495         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
30497 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
30499         PR target/81821
30500         * config/rx/rx.md (BW): New mode attribute.
30501         (sync_lock_test_and_setsi): Add mode suffix to insn output.
30503 2018-01-11  Richard Biener  <rguenther@suse.de>
30505         PR tree-optimization/83435
30506         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
30507         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
30508         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
30510 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30511             Alan Hayward  <alan.hayward@arm.com>
30512             David Sherwood  <david.sherwood@arm.com>
30514         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
30515         field.
30516         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
30517         (aarch64_print_address_internal): Use it to check for a zero offset.
30519 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30520             Alan Hayward  <alan.hayward@arm.com>
30521             David Sherwood  <david.sherwood@arm.com>
30523         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
30524         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
30525         Return a poly_int64 rather than a HOST_WIDE_INT.
30526         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
30527         rather than a HOST_WIDE_INT.
30528         * config/aarch64/aarch64.h (aarch64_frame): Protect with
30529         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
30530         hard_fp_offset, frame_size, initial_adjust, callee_offset and
30531         final_offset from HOST_WIDE_INT to poly_int64.
30532         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
30533         to_constant when getting the number of units in an Advanced SIMD
30534         mode.
30535         (aarch64_builtin_vectorized_function): Check for a constant number
30536         of units.
30537         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
30538         GET_MODE_SIZE.
30539         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
30540         attribute instead of GET_MODE_NUNITS.
30541         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
30542         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
30543         GET_MODE_SIZE for fixed-size registers.
30544         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
30545         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
30546         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
30547         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
30548         (aarch64_print_operand, aarch64_print_address_internal)
30549         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
30550         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
30551         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
30552         Handle polynomial GET_MODE_SIZE.
30553         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
30554         wider than SImode without modification.
30555         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
30556         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
30557         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
30558         passing and returning SVE modes.
30559         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
30560         rather than GEN_INT.
30561         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
30562         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
30563         (aarch64_allocate_and_probe_stack_space): Likewise.
30564         (aarch64_layout_frame): Cope with polynomial offsets.
30565         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
30566         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
30567         polynomial offsets.
30568         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
30569         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
30570         poly_int64 rather than a HOST_WIDE_INT.
30571         (aarch64_get_separate_components, aarch64_process_components)
30572         (aarch64_expand_prologue, aarch64_expand_epilogue)
30573         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
30574         (aarch64_anchor_offset): New function, split out from...
30575         (aarch64_legitimize_address): ...here.
30576         (aarch64_builtin_vectorization_cost): Handle polynomial
30577         TYPE_VECTOR_SUBPARTS.
30578         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
30579         GET_MODE_NUNITS.
30580         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
30581         number of elements from the PARALLEL rather than the mode.
30582         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
30583         rather than GET_MODE_BITSIZE.
30584         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
30585         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
30586         (aarch64_expand_vec_perm_const_1): Handle polynomial
30587         d->perm.length () and d->perm elements.
30588         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
30589         Apply to_constant to d->perm elements.
30590         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
30591         polynomial CONST_VECTOR_NUNITS.
30592         (aarch64_move_pointer): Take amount as a poly_int64 rather
30593         than an int.
30594         (aarch64_progress_pointer): Avoid temporary variable.
30595         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
30596         the mode attribute instead of GET_MODE.
30598 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30599             Alan Hayward  <alan.hayward@arm.com>
30600             David Sherwood  <david.sherwood@arm.com>
30602         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
30603         x exists before using it.
30604         (aarch64_add_constant_internal): Rename to...
30605         (aarch64_add_offset_1): ...this.  Replace regnum with separate
30606         src and dest rtxes.  Handle the case in which they're different,
30607         including when the offset is zero.  Replace scratchreg with an rtx.
30608         Use 2 additions if there is no spare register into which we can
30609         move a 16-bit constant.
30610         (aarch64_add_constant): Delete.
30611         (aarch64_add_offset): Replace reg with separate src and dest
30612         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
30613         Use aarch64_add_offset_1.
30614         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
30615         an rtx rather than an int.  Take the delta as a poly_int64
30616         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
30617         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
30618         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
30619         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
30620         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
30621         and aarch64_add_sp.
30622         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
30623         aarch64_add_constant.
30625 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30627         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
30628         Use scalar_float_mode.
30630 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30632         * config/aarch64/aarch64-simd.md
30633         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
30634         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
30635         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
30636         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
30637         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
30638         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
30639         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
30640         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
30641         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
30642         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
30644 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30646         PR target/83514
30647         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
30648         targ_options->x_arm_arch_string is non NULL.
30650 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
30652         * config/aarch64/aarch64.h
30653         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
30655 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
30657         PR target/82096
30658         * expmed.c (emit_store_flag_force): Swap if const op0
30659         and change VOIDmode to mode of op0.
30661 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30663         PR rtl-optimization/83761
30664         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
30665         than bytes to mode_for_size.
30667 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30669         PR middle-end/83189
30670         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
30671         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
30672         profile.
30674 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30676         PR middle-end/83575
30677         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
30678         when in layout mode.
30679         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
30680         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
30681         partition fixup.
30683 2018-01-10  Michael Collison  <michael.collison@arm.com>
30685         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30686         * config/aarch64/aarch64-option-extension.def: Add
30687         AARCH64_OPT_EXTENSION of 'fp16fml'.
30688         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30689         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30690         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30691         * config/aarch64/constraints.md (Ui7): New constraint.
30692         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30693         (VFMLA_SEL_W): Ditto.
30694         (f16quad): Ditto.
30695         (f16mac1): Ditto.
30696         (VFMLA16_LOW): New int iterator.
30697         (VFMLA16_HIGH): Ditto.
30698         (UNSPEC_FMLAL): New unspec.
30699         (UNSPEC_FMLSL): Ditto.
30700         (UNSPEC_FMLAL2): Ditto.
30701         (UNSPEC_FMLSL2): Ditto.
30702         (f16mac): New code attribute.
30703         * config/aarch64/aarch64-simd-builtins.def
30704         (aarch64_fmlal_lowv2sf): Ditto.
30705         (aarch64_fmlsl_lowv2sf): Ditto.
30706         (aarch64_fmlalq_lowv4sf): Ditto.
30707         (aarch64_fmlslq_lowv4sf): Ditto.
30708         (aarch64_fmlal_highv2sf): Ditto.
30709         (aarch64_fmlsl_highv2sf): Ditto.
30710         (aarch64_fmlalq_highv4sf): Ditto.
30711         (aarch64_fmlslq_highv4sf): Ditto.
30712         (aarch64_fmlal_lane_lowv2sf): Ditto.
30713         (aarch64_fmlsl_lane_lowv2sf): Ditto.
30714         (aarch64_fmlal_laneq_lowv2sf): Ditto.
30715         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30716         (aarch64_fmlalq_lane_lowv4sf): Ditto.
30717         (aarch64_fmlsl_lane_lowv4sf): Ditto.
30718         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30719         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30720         (aarch64_fmlal_lane_highv2sf): Ditto.
30721         (aarch64_fmlsl_lane_highv2sf): Ditto.
30722         (aarch64_fmlal_laneq_highv2sf): Ditto.
30723         (aarch64_fmlsl_laneq_highv2sf): Ditto.
30724         (aarch64_fmlalq_lane_highv4sf): Ditto.
30725         (aarch64_fmlsl_lane_highv4sf): Ditto.
30726         (aarch64_fmlalq_laneq_highv4sf): Ditto.
30727         (aarch64_fmlsl_laneq_highv4sf): Ditto.
30728         * config/aarch64/aarch64-simd.md:
30729         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30730         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30731         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30732         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30733         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30734         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30735         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30736         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30737         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30738         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30739         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30740         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30741         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30742         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30743         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30744         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30745         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30746         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30747         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30748         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30749         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30750         (vfmlsl_low_u32): Ditto.
30751         (vfmlalq_low_u32): Ditto.
30752         (vfmlslq_low_u32): Ditto.
30753         (vfmlal_high_u32): Ditto.
30754         (vfmlsl_high_u32): Ditto.
30755         (vfmlalq_high_u32): Ditto.
30756         (vfmlslq_high_u32): Ditto.
30757         (vfmlal_lane_low_u32): Ditto.
30758         (vfmlsl_lane_low_u32): Ditto.
30759         (vfmlal_laneq_low_u32): Ditto.
30760         (vfmlsl_laneq_low_u32): Ditto.
30761         (vfmlalq_lane_low_u32): Ditto.
30762         (vfmlslq_lane_low_u32): Ditto.
30763         (vfmlalq_laneq_low_u32): Ditto.
30764         (vfmlslq_laneq_low_u32): Ditto.
30765         (vfmlal_lane_high_u32): Ditto.
30766         (vfmlsl_lane_high_u32): Ditto.
30767         (vfmlal_laneq_high_u32): Ditto.
30768         (vfmlsl_laneq_high_u32): Ditto.
30769         (vfmlalq_lane_high_u32): Ditto.
30770         (vfmlslq_lane_high_u32): Ditto.
30771         (vfmlalq_laneq_high_u32): Ditto.
30772         (vfmlslq_laneq_high_u32): Ditto.
30773         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30774         (AARCH64_FL_FOR_ARCH8_4): New.
30775         (AARCH64_ISA_F16FML): New ISA flag.
30776         (TARGET_F16FML): New feature flag for fp16fml.
30777         (doc/invoke.texi): Document new fp16fml option.
30779 2018-01-10  Michael Collison  <michael.collison@arm.com>
30781         * config/aarch64/aarch64-builtins.c:
30782         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30783         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30784         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30785         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30786         (AARCH64_ISA_SHA3): New ISA flag.
30787         (TARGET_SHA3): New feature flag for sha3.
30788         * config/aarch64/iterators.md (sha512_op): New int attribute.
30789         (CRYPTO_SHA512): New int iterator.
30790         (UNSPEC_SHA512H): New unspec.
30791         (UNSPEC_SHA512H2): Ditto.
30792         (UNSPEC_SHA512SU0): Ditto.
30793         (UNSPEC_SHA512SU1): Ditto.
30794         * config/aarch64/aarch64-simd-builtins.def
30795         (aarch64_crypto_sha512hqv2di): New builtin.
30796         (aarch64_crypto_sha512h2qv2di): Ditto.
30797         (aarch64_crypto_sha512su0qv2di): Ditto.
30798         (aarch64_crypto_sha512su1qv2di): Ditto.
30799         (aarch64_eor3qv8hi): Ditto.
30800         (aarch64_rax1qv2di): Ditto.
30801         (aarch64_xarqv2di): Ditto.
30802         (aarch64_bcaxqv8hi): Ditto.
30803         * config/aarch64/aarch64-simd.md:
30804         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30805         (aarch64_crypto_sha512su0qv2di): Ditto.
30806         (aarch64_crypto_sha512su1qv2di): Ditto.
30807         (aarch64_eor3qv8hi): Ditto.
30808         (aarch64_rax1qv2di): Ditto.
30809         (aarch64_xarqv2di): Ditto.
30810         (aarch64_bcaxqv8hi): Ditto.
30811         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30812         (vsha512h2q_u64): Ditto.
30813         (vsha512su0q_u64): Ditto.
30814         (vsha512su1q_u64): Ditto.
30815         (veor3q_u16): Ditto.
30816         (vrax1q_u64): Ditto.
30817         (vxarq_u64): Ditto.
30818         (vbcaxq_u16): Ditto.
30819         * config/arm/types.md (crypto_sha512): New type attribute.
30820         (crypto_sha3): Ditto.
30821         (doc/invoke.texi): Document new sha3 option.
30823 2018-01-10  Michael Collison  <michael.collison@arm.com>
30825         * config/aarch64/aarch64-builtins.c:
30826         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30827         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30828         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30829         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30830         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30831         (AARCH64_ISA_SM4): New ISA flag.
30832         (TARGET_SM4): New feature flag for sm4.
30833         * config/aarch64/aarch64-simd-builtins.def
30834         (aarch64_sm3ss1qv4si): Ditto.
30835         (aarch64_sm3tt1aq4si): Ditto.
30836         (aarch64_sm3tt1bq4si): Ditto.
30837         (aarch64_sm3tt2aq4si): Ditto.
30838         (aarch64_sm3tt2bq4si): Ditto.
30839         (aarch64_sm3partw1qv4si): Ditto.
30840         (aarch64_sm3partw2qv4si): Ditto.
30841         (aarch64_sm4eqv4si): Ditto.
30842         (aarch64_sm4ekeyqv4si): Ditto.
30843         * config/aarch64/aarch64-simd.md:
30844         (aarch64_sm3ss1qv4si): Ditto.
30845         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30846         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30847         (aarch64_sm4eqv4si): Ditto.
30848         (aarch64_sm4ekeyqv4si): Ditto.
30849         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30850         (sm3part_op): Ditto.
30851         (CRYPTO_SM3TT): Ditto.
30852         (CRYPTO_SM3PART): Ditto.
30853         (UNSPEC_SM3SS1): New unspec.
30854         (UNSPEC_SM3TT1A): Ditto.
30855         (UNSPEC_SM3TT1B): Ditto.
30856         (UNSPEC_SM3TT2A): Ditto.
30857         (UNSPEC_SM3TT2B): Ditto.
30858         (UNSPEC_SM3PARTW1): Ditto.
30859         (UNSPEC_SM3PARTW2): Ditto.
30860         (UNSPEC_SM4E): Ditto.
30861         (UNSPEC_SM4EKEY): Ditto.
30862         * config/aarch64/constraints.md (Ui2): New constraint.
30863         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30864         * config/arm/types.md (crypto_sm3): New type attribute.
30865         (crypto_sm4): Ditto.
30866         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30867         (vsm3tt1aq_u32): Ditto.
30868         (vsm3tt1bq_u32): Ditto.
30869         (vsm3tt2aq_u32): Ditto.
30870         (vsm3tt2bq_u32): Ditto.
30871         (vsm3partw1q_u32): Ditto.
30872         (vsm3partw2q_u32): Ditto.
30873         (vsm4eq_u32): Ditto.
30874         (vsm4ekeyq_u32): Ditto.
30875         (doc/invoke.texi): Document new sm4 option.
30877 2018-01-10  Michael Collison  <michael.collison@arm.com>
30879         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30880         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30881         (AARCH64_FL_FOR_ARCH8_4): New.
30882         (AARCH64_FL_V8_4): New flag.
30883         (doc/invoke.texi): Document new armv8.4-a option.
30885 2018-01-10  Michael Collison  <michael.collison@arm.com>
30887         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30888         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30889         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30890         * config/aarch64/aarch64-option-extension.def: Add
30891         AARCH64_OPT_EXTENSION of 'sha2'.
30892         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30893         (crypto): Disable sha2 and aes if crypto disabled.
30894         (crypto): Enable aes and sha2 if enabled.
30895         (simd): Disable sha2 and aes if simd disabled.
30896         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30897         New flags.
30898         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30899         (TARGET_SHA2): New feature flag for sha2.
30900         (TARGET_AES): New feature flag for aes.
30901         * config/aarch64/aarch64-simd.md:
30902         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30903         conditional on TARGET_AES.
30904         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30905         (aarch64_crypto_sha1hsi): Make pattern conditional
30906         on TARGET_SHA2.
30907         (aarch64_crypto_sha1hv4si): Ditto.
30908         (aarch64_be_crypto_sha1hv4si): Ditto.
30909         (aarch64_crypto_sha1su1v4si): Ditto.
30910         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30911         (aarch64_crypto_sha1su0v4si): Ditto.
30912         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30913         (aarch64_crypto_sha256su0v4si): Ditto.
30914         (aarch64_crypto_sha256su1v4si): Ditto.
30915         (doc/invoke.texi): Document new aes and sha2 options.
30917 2018-01-10  Martin Sebor  <msebor@redhat.com>
30919         PR tree-optimization/83781
30920         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30921         as string arrays.
30923 2018-01-11  Martin Sebor  <msebor@gmail.com>
30924             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30926         PR tree-optimization/83501
30927         PR tree-optimization/81703
30929         * tree-ssa-strlen.c (get_string_cst): Rename...
30930         (get_string_len): ...to this.  Handle global constants.
30931         (handle_char_store): Adjust.
30933 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30934             Jim Wilson  <jimw@sifive.com>
30936         * config/riscv/riscv-protos.h (riscv_output_return): New.
30937         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30938         (riscv_attribute_table, riscv_output_return),
30939         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30940         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30941         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30942         function.
30943         (riscv_expand_prologue): Add early return for naked function.
30944         (riscv_expand_epilogue): Likewise.
30945         (riscv_function_ok_for_sibcall): Return false for naked function.
30946         (riscv_set_current_function): New.
30947         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30948         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30949         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30950         * doc/extend.texi (RISC-V Function Attributes): New.
30952 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30954         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30955         check for 128-bit long double before checking TCmode.
30956         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30957         128-bit long doubles before checking TFmode or TCmode.
30958         (FLOAT128_IBM_P): Likewise.
30960 2018-01-10  Martin Sebor  <msebor@redhat.com>
30962         PR tree-optimization/83671
30963         * builtins.c (c_strlen): Unconditionally return zero for the empty
30964         string.
30965         Use -Warray-bounds for warnings.
30966         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30967         for non-constant array indices with COMPONENT_REF, arrays of
30968         arrays, and pointers to arrays.
30969         (gimple_fold_builtin_strlen): Determine and set length range for
30970         non-constant character arrays.
30972 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30974         PR middle-end/81897
30975         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30976         empty blocks.
30978 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30980         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30982 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30984         PR target/83399
30985         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30986         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30987         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30988         indexed_or_indirect_operand predicate.
30989         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30990         (*vsx_le_perm_load_v8hi): Likewise.
30991         (*vsx_le_perm_load_v16qi): Likewise.
30992         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30993         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30994         (*vsx_le_perm_store_v8hi): Likewise.
30995         (*vsx_le_perm_store_v16qi): Likewise.
30996         (eight unnamed splitters): Likewise.
30998 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
31000         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
31001         * config/rs6000/emmintrin.h: Likewise.
31002         * config/rs6000/mmintrin.h: Likewise.
31003         * config/rs6000/xmmintrin.h: Likewise.
31005 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
31007         PR c++/43486
31008         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
31009         "public_flag".
31010         * tree.c (tree_nop_conversion): Return true for location wrapper
31011         nodes.
31012         (maybe_wrap_with_location): New function.
31013         (selftest::check_strip_nops): New function.
31014         (selftest::test_location_wrappers): New function.
31015         (selftest::tree_c_tests): Call it.
31016         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
31017         (maybe_wrap_with_location): New decl.
31018         (EXPR_LOCATION_WRAPPER_P): New macro.
31019         (location_wrapper_p): New inline function.
31020         (tree_strip_any_location_wrapper): New inline function.
31022 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
31024         PR target/83735
31025         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
31026         stack_realign_offset for the largest alignment of stack slot
31027         actually used.
31028         (ix86_find_max_used_stack_alignment): New function.
31029         (ix86_finalize_stack_frame_flags): Use it.  Set
31030         max_used_stack_alignment if we don't realign stack.
31031         * config/i386/i386.h (machine_function): Add
31032         max_used_stack_alignment.
31034 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
31036         * config/arm/arm.opt (-mbranch-cost): New option.
31037         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
31038         account.
31040 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
31042         PR target/83629
31043         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
31044         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
31046 2018-01-10  Richard Biener  <rguenther@suse.de>
31048         PR debug/83765
31049         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
31050         early out so it also covers the case where we have a non-NULL
31051         origin.
31053 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
31055         PR tree-optimization/83753
31056         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
31057         for non-strided grouped accesses if the number of elements is 1.
31059 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
31061         PR target/81616
31062         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
31063         * i386.h (TARGET_USE_GATHER): Define.
31064         * x86-tune.def (X86_TUNE_USE_GATHER): New.
31066 2018-01-10  Martin Liska  <mliska@suse.cz>
31068         PR bootstrap/82831
31069         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
31070         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
31071         partitioning.
31072         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
31073         CLEANUP_NO_PARTITIONING is not set.
31075 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
31077         * doc/rtl.texi: Remove documentation of (const ...) wrappers
31078         for vectors, as a partial revert of r254296.
31079         * rtl.h (const_vec_p): Delete.
31080         (const_vec_duplicate_p): Don't test for vector CONSTs.
31081         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
31082         * expmed.c (make_tree): Likewise.
31084         Revert:
31085         * common.md (E, F): Use CONSTANT_P instead of checking for
31086         CONST_VECTOR.
31087         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
31088         checking for CONST_VECTOR.
31090 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31092         PR middle-end/83575
31093         * predict.c (force_edge_cold): Handle in more sane way edges
31094         with no prediction.
31096 2018-01-09  Carl Love  <cel@us.ibm.com>
31098         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
31099         V4SI, V4SF types.
31100         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
31101         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
31102         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
31103         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
31104         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
31105         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
31106         * config/rs6000/rs6000-protos.h: Add extern defition for
31107         rs6000_generate_float2_double_code.
31108         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
31109         function.
31110         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
31111         (float2_v2df): Add define_expand.
31113 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
31115         PR target/83628
31116         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
31117         op_mode in the force_to_mode call.
31119 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31121         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
31122         instead of checking each element individually.
31123         (aarch64_evpc_uzp): Likewise.
31124         (aarch64_evpc_zip): Likewise.
31125         (aarch64_evpc_ext): Likewise.
31126         (aarch64_evpc_rev): Likewise.
31127         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
31128         instead of checking each element individually.  Return true without
31129         generating rtl if
31130         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
31131         whether all selected elements come from the same input, instead of
31132         checking each element individually.  Remove calls to gen_rtx_REG,
31133         start_sequence and end_sequence and instead assert that no rtl is
31134         generated.
31136 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31138         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
31139         order of HIGH and CONST checks.
31141 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31143         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
31144         if the destination isn't an SSA_NAME.
31146 2018-01-09  Richard Biener  <rguenther@suse.de>
31148         PR tree-optimization/83668
31149         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
31150         move prologue...
31151         (canonicalize_loop_form): ... here, renamed from ...
31152         (canonicalize_loop_closed_ssa_form): ... this and amended to
31153         swap successor edges for loop exit blocks to make us use
31154         the RPO order we need for initial schedule generation.
31156 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
31158         PR tree-optimization/64811
31159         * match.pd: When optimizing comparisons with Inf, avoid
31160         introducing or losing exceptions from comparisons with NaN.
31162 2018-01-09  Martin Liska  <mliska@suse.cz>
31164         PR sanitizer/82517
31165         * asan.c (shadow_mem_size): Add gcc_assert.
31167 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
31169         Don't save registers in main().
31171         PR target/83738
31172         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
31173         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
31174         * config/avr/avr.c (avr_set_current_function): Don't error if
31175         naked, OS_task or OS_main are specified at the same time.
31176         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
31177         OS_main.
31178         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
31179         attribute.
31180         * common/config/avr/avr-common.c (avr_option_optimization_table):
31181         Switch on -mmain-is-OS_task for optimizing compilations.
31183 2018-01-09  Richard Biener  <rguenther@suse.de>
31185         PR tree-optimization/83572
31186         * graphite.c: Include cfganal.h.
31187         (graphite_transform_loops): Connect infinite loops to exit
31188         and remove fake edges at the end.
31190 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31192         * ipa-inline.c (edge_badness): Revert accidental checkin.
31194 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31196         PR ipa/80763
31197         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
31198         symbols; not inline clones.
31200 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
31202         PR target/83507
31203         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
31204         hard registers.  Formatting fixes.
31206         PR preprocessor/83722
31207         * gcc.c (try_generate_repro): Pass
31208         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
31209         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
31210         do_report_bug.
31212 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
31213             Kito Cheng  <kito.cheng@gmail.com>
31215         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
31216         (riscv_leaf_function_p): Delete.
31217         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
31219 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31221         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
31222         function.
31223         (do_ifelse): New function.
31224         (do_isel): New function.
31225         (do_sub3): New function.
31226         (do_add3): New function.
31227         (do_load_mask_compare): New function.
31228         (do_overlap_load_compare): New function.
31229         (expand_compare_loop): New function.
31230         (expand_block_compare): Call expand_compare_loop() when appropriate.
31231         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
31232         option description.
31233         (-mblock-compare-inline-loop-limit): New option.
31235 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31237         PR target/83677
31238         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
31239         Reverse order of second and third operands in first alternative.
31240         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
31241         of first and second elements in UNSPEC_VPERMR vector.
31242         (altivec_expand_vec_perm_le): Likewise.
31244 2018-01-08  Jeff Law  <law@redhat.com>
31246         PR rtl-optimizatin/81308
31247         * tree-switch-conversion.c (cfg_altered): New file scoped static.
31248         (process_switch): If group_case_labels makes a change, then set
31249         cfg_altered.
31250         (pass_convert_switch::execute): If a switch is converted, then
31251         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
31253         PR rtl-optimization/81308
31254         * recog.c (split_all_insns): Conditionally cleanup the CFG after
31255         splitting insns.
31257 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
31259         PR target/83663 - Revert r255946
31260         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
31261         generation for cases where splatting a value is not useful.
31262         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
31263         across a vec_duplicate and a paradoxical subreg forming a vector
31264         mode to a vec_concat.
31266 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31268         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
31269         -march=armv8.3-a variants.
31270         * config/arm/t-multilib: Likewise.
31271         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
31273 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31275         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
31276         to generate rtl.
31277         (cceq_ior_compare_complement): Give it a name so I can use it, and
31278         change boolean_or_operator predicate to boolean_operator so it can
31279         be used to generate a crand.
31280         (eqne): New code iterator.
31281         (bd/bd_neg): New code_attrs.
31282         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
31283         a single define_insn.
31284         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
31285         decrement (bdnzt/bdnzf/bdzt/bdzf).
31286         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
31287         with the new names of the branch decrement patterns, and added the
31288         names of the branch decrement conditional patterns.
31290 2018-01-08  Richard Biener  <rguenther@suse.de>
31292         PR tree-optimization/83563
31293         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
31294         cache.
31296 2018-01-08  Richard Biener  <rguenther@suse.de>
31298         PR middle-end/83713
31299         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
31301 2018-01-08  Richard Biener  <rguenther@suse.de>
31303         PR tree-optimization/83685
31304         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
31305         references to abnormals.
31307 2018-01-08  Richard Biener  <rguenther@suse.de>
31309         PR lto/83719
31310         * dwarf2out.c (output_indirect_strings): Handle empty
31311         skeleton_debug_str_hash.
31312         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
31314 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
31316         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
31317         (emit_store_direct): Likewise.
31318         (arc_trampoline_adjust_address): Likewise.
31319         (arc_asm_trampoline_template): New function.
31320         (arc_initialize_trampoline): Use asm_trampoline_template.
31321         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
31322         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
31323         * config/arc/arc.md (flush_icache): Delete pattern.
31325 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
31327         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
31328         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
31329         munaligned-access.
31331 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31333         PR target/83681
31334         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
31335         by not USED_FOR_TARGET.
31336         (make_pass_resolve_sw_modes): Likewise.
31338 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31340         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
31341         USED_FOR_TARGET.
31343 2018-01-08  Richard Biener  <rguenther@suse.de>
31345         PR middle-end/83580
31346         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
31348 2018-01-08  Richard Biener  <rguenther@suse.de>
31350         PR middle-end/83517
31351         * match.pd ((t * 2) / 2) -> t): Add missing :c.
31353 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
31355         PR middle-end/81897
31356         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
31357         basic blocks with a small number of successors.
31358         (convert_control_dep_chain_into_preds): Improve handling of
31359         forwarder blocks.
31360         (dump_predicates): Split apart into...
31361         (dump_pred_chain): ...here...
31362         (dump_pred_info): ...and here.
31363         (can_one_predicate_be_invalidated_p): Add debugging printfs.
31364         (can_chain_union_be_invalidated_p): Improve check for invalidation
31365         of paths.
31366         (uninit_uses_cannot_happen): Avoid unnecessary if
31367         convert_control_dep_chain_into_preds yielded nothing.
31369 2018-01-06  Martin Sebor  <msebor@redhat.com>
31371         PR tree-optimization/83640
31372         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
31373         subtracting negative offset from size.
31374         (builtin_access::overlap): Adjust offset bounds of the access to fall
31375         within the size of the object if possible.
31377 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
31379         PR rtl-optimization/83699
31380         * expmed.c (extract_bit_field_1): Restrict the vector usage of
31381         extract_bit_field_as_subreg to cases in which the extracted
31382         value is also a vector.
31384         * lra-constraints.c (process_alt_operands): Test for the equivalence
31385         substitutions when detecting a possible reload cycle.
31387 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
31389         PR debug/83480
31390         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
31391         by default if flag_selective_schedling{,2}.  Formatting fixes.
31393         PR rtl-optimization/83682
31394         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
31395         if it has non-VECTOR_MODE element mode.
31396         (vec_duplicate_p): Likewise.
31398         PR middle-end/83694
31399         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
31400         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
31402 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
31404         PR target/83604
31405         * config/i386/i386-builtin.def
31406         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
31407         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
31408         Require also OPTION_MASK_ISA_AVX512F in addition to
31409         OPTION_MASK_ISA_GFNI.
31410         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
31411         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
31412         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
31413         to OPTION_MASK_ISA_GFNI.
31414         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
31415         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
31416         OPTION_MASK_ISA_AVX512BW.
31417         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
31418         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
31419         addition to OPTION_MASK_ISA_GFNI.
31420         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
31421         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
31422         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
31423         to OPTION_MASK_ISA_GFNI.
31424         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
31425         a requirement for all ISAs rather than any of them with a few
31426         exceptions.
31427         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
31428         processing.
31429         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
31430         bitmasks to be enabled with 3 exceptions, instead of requiring any
31431         enabled ISA with lots of exceptions.
31432         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
31433         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
31434         Change avx512bw in isa attribute to avx512f.
31435         * config/i386/sgxintrin.h: Add license boilerplate.
31436         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
31437         to __AVX512F__ and __AVX512VL to __AVX512VL__.
31438         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
31439         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
31440         defined.
31441         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
31442         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
31443         temporarily sse2 rather than sse if not enabled already.
31445         PR target/83604
31446         * config/i386/sse.md (VI248_VLBW): Rename to ...
31447         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
31448         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
31449         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
31450         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
31451         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
31452         mode iterator instead of VI248_VLBW.
31454 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
31456         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
31457         (record_modified): Skip clobbers; add debug output.
31458         (param_change_prob): Use sreal frequencies.
31460 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
31462         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
31463         punt for user-aligned variables.
31465 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
31467         * tree-chrec.c (chrec_contains_symbols): Return true for
31468         POLY_INT_CST.
31470 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
31472         PR target/82439
31473         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
31474         of (x|y) == x for BICS pattern.
31476 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
31478         PR tree-optimization/83605
31479         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
31480         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
31481         can throw.
31483 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31485         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
31486         * config/epiphany/rtems.h: New file.
31488 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
31489             Uros Bizjak  <ubizjak@gmail.com>
31491         PR target/83554
31492         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
31493         QIreg_operand instead of register_operand predicate.
31494         * config/i386/i386.c (ix86_rop_should_change_byte_p,
31495         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
31496         comments instead of -fmitigate[-_]rop.
31498 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
31500         PR bootstrap/81926
31501         * cgraphunit.c (symbol_table::compile): Switch to text_section
31502         before calling assembly_start debug hook.
31503         * run-rtl-passes.c (run_rtl_passes): Likewise.
31504         Include output.h.
31506 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31508         * tree-vrp.c (extract_range_from_binary_expr_1): Check
31509         range_int_cst_p rather than !symbolic_range_p before calling
31510         extract_range_from_multiplicative_op_1.
31512 2018-01-04  Jeff Law  <law@redhat.com>
31514         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
31515         redundant test in assertion.
31517 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31519         * doc/rtl.texi: Document machine_mode wrapper classes.
31521 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31523         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
31524         using tree_to_uhwi.
31526 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31528         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
31529         the VEC_PERM_EXPR fold to fail.
31531 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
31533         PR debug/83585
31534         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
31535         to switched_sections.
31537 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31539         PR target/83680
31540         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
31541         test for d.testing.
31543 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
31545         PR target/83387
31546         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
31547         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
31549 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
31551         PR debug/83666
31552         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
31553         is BLKmode and bitpos not zero or mode change is needed.
31555 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31557         PR target/83675
31558         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
31559         TARGET_VIS2.
31561 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
31563         PR target/83628
31564         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
31565         instead of MULT rtx.  Update all corresponding splitters.
31566         (*saddl_se): Ditto.
31567         (*ssub<modesuffix>): Ditto.
31568         (*ssubl_se): Ditto.
31569         (*cmp_sadd_di): Update split patterns.
31570         (*cmp_sadd_si): Ditto.
31571         (*cmp_sadd_sidi): Ditto.
31572         (*cmp_ssub_di): Ditto.
31573         (*cmp_ssub_si): Ditto.
31574         (*cmp_ssub_sidi): Ditto.
31575         * config/alpha/predicates.md (const23_operand): New predicate.
31576         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
31577         Look for ASHIFT, not MULT inner operand.
31578         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
31580 2018-01-04  Martin Liska  <mliska@suse.cz>
31582         PR gcov-profile/83669
31583         * gcov.c (output_intermediate_file): Add version to intermediate
31584         gcov file.
31585         * doc/gcov.texi: Document new field 'version' in intermediate
31586         file format. Fix location of '-k' option of gcov command.
31588 2018-01-04  Martin Liska  <mliska@suse.cz>
31590         PR ipa/82352
31591         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
31593 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
31595         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
31597 2018-01-03  Martin Sebor  <msebor@redhat.com>
31599         PR tree-optimization/83655
31600         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
31601         checking calls with invalid arguments.
31603 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31605         * tree-vect-stmts.c (vect_get_store_rhs): New function.
31606         (vectorizable_mask_load_store): Delete.
31607         (vectorizable_call): Return false for masked loads and stores.
31608         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
31609         instead of gimple_assign_rhs1.
31610         (vectorizable_load): Handle IFN_MASK_LOAD.
31611         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
31613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31615         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
31616         split out from..,
31617         (vectorizable_mask_load_store): ...here.
31618         (vectorizable_load): ...and here.
31620 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31622         * tree-vect-stmts.c (vect_build_all_ones_mask)
31623         (vect_build_zero_merge_argument): New functions, split out from...
31624         (vectorizable_load): ...here.
31626 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31628         * tree-vect-stmts.c (vect_check_store_rhs): New function,
31629         split out from...
31630         (vectorizable_mask_load_store): ...here.
31631         (vectorizable_store): ...and here.
31633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31635         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
31636         split out from...
31637         (vectorizable_mask_load_store): ...here.
31639 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31641         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
31642         (vect_model_store_cost): Take a vec_load_store_type instead of a
31643         vect_def_type.
31644         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
31645         (vect_model_store_cost): Take a vec_load_store_type instead of a
31646         vect_def_type.
31647         (vectorizable_mask_load_store): Update accordingly.
31648         (vectorizable_store): Likewise.
31649         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
31651 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31653         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
31654         IFN_MASK_LOAD calls here rather than...
31655         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
31657 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31658             Alan Hayward  <alan.hayward@arm.com>
31659             David Sherwood  <david.sherwood@arm.com>
31661         * expmed.c (extract_bit_field_1): For vector extracts,
31662         fall back to extract_bit_field_as_subreg if vec_extract
31663         isn't available.
31665 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31666             Alan Hayward  <alan.hayward@arm.com>
31667             David Sherwood  <david.sherwood@arm.com>
31669         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
31670         they are variable or constant sized.
31671         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
31672         slots for constant-sized data.
31674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31675             Alan Hayward  <alan.hayward@arm.com>
31676             David Sherwood  <david.sherwood@arm.com>
31678         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
31679         handling COND_EXPRs with boolean comparisons, try to find a better
31680         basis for the mask type than the boolean itself.
31682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31684         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31685         is calculated and how it can be overridden.
31686         * genmodes.c (max_bitsize_mode_any_mode): New variable.
31687         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31688         if defined.
31689         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31690         if nonzero.
31692 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31693             Alan Hayward  <alan.hayward@arm.com>
31694             David Sherwood  <david.sherwood@arm.com>
31696         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31697         Remove the mode argument.
31698         (aarch64_simd_valid_immediate): Remove the mode and inverse
31699         arguments.
31700         * config/aarch64/iterators.md (bitsize): New iterator.
31701         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31702         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31703         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31704         aarch64_simd_valid_immediate.
31705         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31706         (aarch64_reg_or_bic_imm): Likewise.
31707         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31708         with an insn_type enum and msl with a modifier_type enum.
31709         Replace element_width with a scalar_mode.  Change the shift
31710         to unsigned int.  Add constructors for scalar_float_mode and
31711         scalar_int_mode elements.
31712         (aarch64_vect_float_const_representable_p): Delete.
31713         (aarch64_can_const_movi_rtx_p)
31714         (aarch64_simd_scalar_immediate_valid_for_move)
31715         (aarch64_simd_make_constant): Update call to
31716         aarch64_simd_valid_immediate.
31717         (aarch64_advsimd_valid_immediate_hs): New function.
31718         (aarch64_advsimd_valid_immediate): Likewise.
31719         (aarch64_simd_valid_immediate): Remove mode and inverse
31720         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
31721         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31722         and aarch64_float_const_representable_p on the result.
31723         (aarch64_output_simd_mov_immediate): Remove mode argument.
31724         Update call to aarch64_simd_valid_immediate and use of
31725         simd_immediate_info.
31726         (aarch64_output_scalar_simd_mov_immediate): Update call
31727         accordingly.
31729 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31730             Alan Hayward  <alan.hayward@arm.com>
31731             David Sherwood  <david.sherwood@arm.com>
31733         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31734         (mode_nunits): Likewise CONST_MODE_NUNITS.
31735         * machmode.def (ADJUST_NUNITS): Document.
31736         * genmodes.c (mode_data::need_nunits_adj): New field.
31737         (blank_mode): Update accordingly.
31738         (adj_nunits): New variable.
31739         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31740         parameter.
31741         (emit_mode_size_inline): Set need_bytesize_adj for all modes
31742         listed in adj_nunits.
31743         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31744         listed in adj_nunits.  Don't emit case statements for such modes.
31745         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31746         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
31747         nothing if adj_nunits is nonnull.
31748         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31749         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31750         (emit_mode_fbit): Update use of print_maybe_const_decl.
31751         (emit_move_size): Likewise.  Treat the array as non-const
31752         if adj_nunits.
31753         (emit_mode_adjustments): Handle adj_nunits.
31755 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31757         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31758         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31759         (VECTOR_MODES): Use it.
31760         (make_vector_modes): Take the prefix as an argument.
31762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31763             Alan Hayward  <alan.hayward@arm.com>
31764             David Sherwood  <david.sherwood@arm.com>
31766         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31767         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31768         for MODE_VECTOR_BOOL.
31769         * machmode.def (VECTOR_BOOL_MODE): Document.
31770         * genmodes.c (VECTOR_BOOL_MODE): New macro.
31771         (make_vector_bool_mode): New function.
31772         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31773         MODE_VECTOR_BOOL.
31774         * lto-streamer-in.c (lto_input_mode_table): Likewise.
31775         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31776         Likewise.
31777         * stor-layout.c (int_mode_for_mode): Likewise.
31778         * tree.c (build_vector_type_for_mode): Likewise.
31779         * varasm.c (output_constant_pool_2): Likewise.
31780         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31781         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
31782         for MODE_VECTOR_BOOL.
31783         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31784         of mode class checks.
31785         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31786         instead of a list of mode class checks.
31787         (expand_vector_scalar_condition): Likewise.
31788         (type_for_widest_vector_mode): Handle BImode as an inner mode.
31790 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31791             Alan Hayward  <alan.hayward@arm.com>
31792             David Sherwood  <david.sherwood@arm.com>
31794         * machmode.h (mode_size): Change from unsigned short to
31795         poly_uint16_pod.
31796         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31797         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31798         or if measurement_type is not polynomial.
31799         (fixed_size_mode::includes_p): Check for constant-sized modes.
31800         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31801         return a poly_uint16 rather than an unsigned short.
31802         (emit_mode_size): Change the type of mode_size from unsigned short
31803         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
31804         (emit_mode_adjustments): Cope with polynomial vector sizes.
31805         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31806         for GET_MODE_SIZE.
31807         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31808         for GET_MODE_SIZE.
31809         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31810         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31811         * caller-save.c (setup_save_areas): Likewise.
31812         (replace_reg_with_saved_mem): Likewise.
31813         * calls.c (emit_library_call_value_1): Likewise.
31814         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31815         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31816         (gen_lowpart_for_combine): Likewise.
31817         * convert.c (convert_to_integer_1): Likewise.
31818         * cse.c (equiv_constant, cse_insn): Likewise.
31819         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31820         (cselib_subst_to_values): Likewise.
31821         * dce.c (word_dce_process_block): Likewise.
31822         * df-problems.c (df_word_lr_mark_ref): Likewise.
31823         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31824         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31825         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31826         (rtl_for_decl_location): Likewise.
31827         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31828         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31829         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31830         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31831         (expand_expr_real_1): Likewise.
31832         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31833         (pad_below): Likewise.
31834         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31835         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31836         * ira.c (get_subreg_tracking_sizes): Likewise.
31837         * ira-build.c (ira_create_allocno_objects): Likewise.
31838         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31839         (ira_sort_regnos_for_alter_reg): Likewise.
31840         * ira-costs.c (record_operand_costs): Likewise.
31841         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31842         (resolve_simple_move): Likewise.
31843         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31844         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31845         (lra_constraints): Likewise.
31846         (CONST_POOL_OK_P): Reject variable-sized modes.
31847         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31848         (add_pseudo_to_slot, lra_spill): Likewise.
31849         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31850         * optabs-query.c (get_best_extraction_insn): Likewise.
31851         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31852         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31853         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31854         * recog.c (offsettable_address_addr_space_p): Likewise.
31855         * regcprop.c (maybe_mode_change): Likewise.
31856         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31857         * regrename.c (build_def_use): Likewise.
31858         * regstat.c (dump_reg_info): Likewise.
31859         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31860         (find_reloads, find_reloads_subreg_address): Likewise.
31861         * reload1.c (eliminate_regs_1): Likewise.
31862         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31863         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31864         (simplify_binary_operation_1, simplify_subreg): Likewise.
31865         * targhooks.c (default_function_arg_padding): Likewise.
31866         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31867         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31868         (verify_gimple_assign_ternary): Likewise.
31869         * tree-inline.c (estimate_move_cost): Likewise.
31870         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31871         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31872         (get_address_cost_ainc): Likewise.
31873         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31874         (vect_supportable_dr_alignment): Likewise.
31875         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31876         (vectorizable_reduction): Likewise.
31877         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31878         (vectorizable_operation, vectorizable_load): Likewise.
31879         * tree.c (build_same_sized_truth_vector_type): Likewise.
31880         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31881         * var-tracking.c (emit_note_insn_var_location): Likewise.
31882         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31883         (ADDR_VEC_ALIGN): Likewise.
31885 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31886             Alan Hayward  <alan.hayward@arm.com>
31887             David Sherwood  <david.sherwood@arm.com>
31889         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31890         unsigned short.
31891         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31892         or if measurement_type is polynomial.
31893         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31894         * combine.c (make_extraction): Likewise.
31895         * dse.c (find_shift_sequence): Likewise.
31896         * dwarf2out.c (mem_loc_descriptor): Likewise.
31897         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31898         (extract_bit_field, extract_low_bits): Likewise.
31899         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31900         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31901         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31902         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31903         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31904         * reload.c (find_reloads): Likewise.
31905         * reload1.c (alter_reg): Likewise.
31906         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31907         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31908         * tree-if-conv.c (predicate_mem_writes): Likewise.
31909         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31910         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31911         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31912         * valtrack.c (dead_debug_insert_temp): Likewise.
31913         * varasm.c (mergeable_constant_section): Likewise.
31914         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31916 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31917             Alan Hayward  <alan.hayward@arm.com>
31918             David Sherwood  <david.sherwood@arm.com>
31920         * expr.c (expand_assignment): Cope with polynomial mode sizes
31921         when assigning to a CONCAT.
31923 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31924             Alan Hayward  <alan.hayward@arm.com>
31925             David Sherwood  <david.sherwood@arm.com>
31927         * machmode.h (mode_precision): Change from unsigned short to
31928         poly_uint16_pod.
31929         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31930         short.
31931         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31932         or if measurement_type is not polynomial.
31933         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31934         in which the mode is already known to be a scalar_int_mode.
31935         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31936         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31937         initializer.
31938         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31939         for GET_MODE_PRECISION.
31940         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31941         for GET_MODE_PRECISION.
31942         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31943         as polynomial.
31944         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31945         (expand_field_assignment, make_extraction): Likewise.
31946         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31947         (get_last_value): Likewise.
31948         * convert.c (convert_to_integer_1): Likewise.
31949         * cse.c (cse_insn): Likewise.
31950         * expr.c (expand_expr_real_1): Likewise.
31951         * lra-constraints.c (simplify_operand_subreg): Likewise.
31952         * optabs-query.c (can_atomic_load_p): Likewise.
31953         * optabs.c (expand_atomic_load): Likewise.
31954         (expand_atomic_store): Likewise.
31955         * ree.c (combine_reaching_defs): Likewise.
31956         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31957         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31958         * tree.h (type_has_mode_precision_p): Likewise.
31959         * ubsan.c (instrument_si_overflow): Likewise.
31961 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31962             Alan Hayward  <alan.hayward@arm.com>
31963             David Sherwood  <david.sherwood@arm.com>
31965         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31966         polynomial numbers of units.
31967         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31968         (valid_vector_subparts_p): New function.
31969         (build_vector_type): Remove temporary shim and take the number
31970         of units as a poly_uint64 rather than an int.
31971         (build_opaque_vector_type): Take the number of units as a
31972         poly_uint64 rather than an int.
31973         * tree.c (build_vector_from_ctor): Handle polynomial
31974         TYPE_VECTOR_SUBPARTS.
31975         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31976         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31977         (build_vector_from_val): If the number of units is variable,
31978         use build_vec_duplicate_cst for constant operands and
31979         VEC_DUPLICATE_EXPR otherwise.
31980         (make_vector_type): Remove temporary is_constant ().
31981         (build_vector_type, build_opaque_vector_type): Take the number of
31982         units as a poly_uint64 rather than an int.
31983         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31984         VECTOR_CST_NELTS.
31985         * cfgexpand.c (expand_debug_expr): Likewise.
31986         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31987         (store_constructor, expand_expr_real_1): Likewise.
31988         (const_scalar_mask_from_tree): Likewise.
31989         * fold-const-call.c (fold_const_reduction): Likewise.
31990         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31991         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31992         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31993         (fold_relational_const): Likewise.
31994         (native_interpret_vector): Likewise.  Change the size from an
31995         int to an unsigned int.
31996         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31997         TYPE_VECTOR_SUBPARTS.
31998         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31999         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
32000         duplicating a non-constant operand into a variable-length vector.
32001         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
32002         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
32003         * ipa-icf.c (sem_variable::equals): Likewise.
32004         * match.pd: Likewise.
32005         * omp-simd-clone.c (simd_clone_subparts): Likewise.
32006         * print-tree.c (print_node): Likewise.
32007         * stor-layout.c (layout_type): Likewise.
32008         * targhooks.c (default_builtin_vectorization_cost): Likewise.
32009         * tree-cfg.c (verify_gimple_comparison): Likewise.
32010         (verify_gimple_assign_binary): Likewise.
32011         (verify_gimple_assign_ternary): Likewise.
32012         (verify_gimple_assign_single): Likewise.
32013         * tree-pretty-print.c (dump_generic_node): Likewise.
32014         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32015         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
32016         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
32017         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
32018         (vect_shift_permute_load_chain): Likewise.
32019         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
32020         (expand_vector_condition, optimize_vector_constructor): Likewise.
32021         (lower_vec_perm, get_compute_type): Likewise.
32022         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
32023         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
32024         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
32025         (vect_recog_mask_conversion_pattern): Likewise.
32026         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
32027         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
32028         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32029         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
32030         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
32031         (vectorizable_shift, vectorizable_operation, vectorizable_store)
32032         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
32033         (supportable_widening_operation): Likewise.
32034         (supportable_narrowing_operation): Likewise.
32035         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
32036         Likewise.
32037         * varasm.c (output_constant): Likewise.
32039 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32040             Alan Hayward  <alan.hayward@arm.com>
32041             David Sherwood  <david.sherwood@arm.com>
32043         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
32044         so that both the length == 3 and length != 3 cases set up their
32045         own permute vectors.  Add comments explaining why we know the
32046         number of elements is constant.
32047         (vect_permute_load_chain): Likewise.
32049 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32050             Alan Hayward  <alan.hayward@arm.com>
32051             David Sherwood  <david.sherwood@arm.com>
32053         * machmode.h (mode_nunits): Change from unsigned char to
32054         poly_uint16_pod.
32055         (ONLY_FIXED_SIZE_MODES): New macro.
32056         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
32057         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
32058         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
32059         New typedefs.
32060         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
32061         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
32062         or if measurement_type is not polynomial.
32063         * genmodes.c (ZERO_COEFFS): New macro.
32064         (emit_mode_nunits_inline): Make mode_nunits_inline return a
32065         poly_uint16.
32066         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
32067         Use ZERO_COEFFS when emitting initializers.
32068         * data-streamer.h (bp_pack_poly_value): New function.
32069         (bp_unpack_poly_value): Likewise.
32070         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32071         for GET_MODE_NUNITS.
32072         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32073         for GET_MODE_NUNITS.
32074         * tree.c (make_vector_type): Remove temporary shim and make
32075         the real function take the number of units as a poly_uint64
32076         rather than an int.
32077         (build_vector_type_for_mode): Handle polynomial nunits.
32078         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
32079         * emit-rtl.c (const_vec_series_p_1): Likewise.
32080         (gen_rtx_CONST_VECTOR): Likewise.
32081         * fold-const.c (test_vec_duplicate_folding): Likewise.
32082         * genrecog.c (validate_pattern): Likewise.
32083         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
32084         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
32085         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
32086         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
32087         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
32088         * rtlanal.c (subreg_get_info): Likewise.
32089         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32090         (vect_grouped_load_supported): Likewise.
32091         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
32092         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
32093         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
32094         (simplify_const_unary_operation, simplify_binary_operation_1)
32095         (simplify_const_binary_operation, simplify_ternary_operation)
32096         (test_vector_ops_duplicate, test_vector_ops): Likewise.
32097         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
32098         instead of CONST_VECTOR_NUNITS.
32099         * varasm.c (output_constant_pool_2): Likewise.
32100         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
32101         explicit-encoded elements in the XVEC for variable-length vectors.
32103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32105         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
32107 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32108             Alan Hayward  <alan.hayward@arm.com>
32109             David Sherwood  <david.sherwood@arm.com>
32111         * coretypes.h (fixed_size_mode): Declare.
32112         (fixed_size_mode_pod): New typedef.
32113         * builtins.h (target_builtins::x_apply_args_mode)
32114         (target_builtins::x_apply_result_mode): Change type to
32115         fixed_size_mode_pod.
32116         * builtins.c (apply_args_size, apply_result_size, result_vector)
32117         (expand_builtin_apply_args_1, expand_builtin_apply)
32118         (expand_builtin_return): Update accordingly.
32120 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32122         * cse.c (hash_rtx_cb): Hash only the encoded elements.
32123         * cselib.c (cselib_hash_rtx): Likewise.
32124         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
32125         CONST_VECTOR encoding.
32127 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32128             Jeff Law  <law@redhat.com>
32130         PR target/83641
32131         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
32132         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
32133         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
32134         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
32136         PR target/83641
32137         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
32138         explicitly probe *sp in a noreturn function if there were any callee
32139         register saves or frame pointer is needed.
32141 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32143         PR debug/83621
32144         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
32145         BLKmode for ternary, binary or unary expressions.
32147         PR debug/83645
32148         * var-tracking.c (delete_vta_debug_insn): New inline function.
32149         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
32150         insns from get_insns () to NULL instead of each bb separately.
32151         Use delete_vta_debug_insn.  No longer static.
32152         (vt_debug_insns_local, variable_tracking_main_1): Adjust
32153         delete_vta_debug_insns callers.
32154         * rtl.h (delete_vta_debug_insns): Declare.
32155         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
32156         instead of variable_tracking_main.
32158 2018-01-03  Martin Sebor  <msebor@redhat.com>
32160         PR tree-optimization/83603
32161         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
32162         arguments past the endof the argument list in functions declared
32163         without a prototype.
32164         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
32165         Avoid checking when arguments are null.
32167 2018-01-03  Martin Sebor  <msebor@redhat.com>
32169         PR c/83559
32170         * doc/extend.texi (attribute const): Fix a typo.
32171         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
32172         issuing -Wsuggest-attribute for void functions.
32174 2018-01-03  Martin Sebor  <msebor@redhat.com>
32176         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
32177         offset_int::from instead of wide_int::to_shwi.
32178         (maybe_diag_overlap): Remove assertion.
32179         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
32180         * gimple-ssa-sprintf.c (format_directive): Same.
32181         (parse_directive): Same.
32182         (sprintf_dom_walker::compute_format_length): Same.
32183         (try_substitute_return_value): Same.
32185 2018-01-03  Jeff Law  <law@redhat.com>
32187         PR middle-end/83654
32188         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
32189         non-constant residual for zero at runtime and avoid probing in
32190         that case.  Reorganize code for trailing problem to mirror handling
32191         of the residual.
32193 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32195         PR tree-optimization/83501
32196         * tree-ssa-strlen.c (get_string_cst): New.
32197         (handle_char_store): Call get_string_cst.
32199 2018-01-03  Martin Liska  <mliska@suse.cz>
32201         PR tree-optimization/83593
32202         * tree-ssa-strlen.c: Include tree-cfg.h.
32203         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
32204         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
32205         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
32206         to false.
32207         (strlen_dom_walker::before_dom_children): Call
32208         gimple_purge_dead_eh_edges. Dump tranformation with details
32209         dump flags.
32210         (strlen_dom_walker::before_dom_children): Update call by adding
32211         new argument cleanup_eh.
32212         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
32214 2018-01-03  Martin Liska  <mliska@suse.cz>
32216         PR ipa/83549
32217         * cif-code.def (VARIADIC_THUNK): New enum value.
32218         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
32219         thunks.
32221 2018-01-03  Jan Beulich  <jbeulich@suse.com>
32223         * sse.md (mov<mode>_internal): Tighten condition for when to use
32224         vmovdqu<ssescalarsize> for TI and OI modes.
32226 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32228         Update copyright years.
32230 2018-01-03  Martin Liska  <mliska@suse.cz>
32232         PR ipa/83594
32233         * ipa-visibility.c (function_and_variable_visibility): Skip
32234         functions with noipa attribure.
32236 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32238         * gcc.c (process_command): Update copyright notice dates.
32239         * gcov-dump.c (print_version): Ditto.
32240         * gcov.c (print_version): Ditto.
32241         * gcov-tool.c (print_version): Ditto.
32242         * gengtype.c (create_file): Ditto.
32243         * doc/cpp.texi: Bump @copying's copyright year.
32244         * doc/cppinternals.texi: Ditto.
32245         * doc/gcc.texi: Ditto.
32246         * doc/gccint.texi: Ditto.
32247         * doc/gcov.texi: Ditto.
32248         * doc/install.texi: Ditto.
32249         * doc/invoke.texi: Ditto.
32251 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32253         * vector-builder.h (vector_builder::m_full_nelts): Change from
32254         unsigned int to poly_uint64.
32255         (vector_builder::full_nelts): Update prototype accordingly.
32256         (vector_builder::new_vector): Likewise.
32257         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
32258         (vector_builder::operator ==): Likewise.
32259         (vector_builder::finalize): Likewise.
32260         * int-vector-builder.h (int_vector_builder::int_vector_builder):
32261         Take the number of elements as a poly_uint64 rather than an
32262         unsigned int.
32263         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
32264         from unsigned int to poly_uint64.
32265         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
32266         (vec_perm_indices::new_vector): Likewise.
32267         (vec_perm_indices::length): Likewise.
32268         (vec_perm_indices::nelts_per_input): Likewise.
32269         (vec_perm_indices::input_nelts): Likewise.
32270         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
32271         number of elements per input as a poly_uint64 rather than an
32272         unsigned int.  Use the original encoding for variable-length
32273         vectors, rather than clamping each individual element.
32274         For the second and subsequent elements in each pattern,
32275         clamp the step and base before clamping their sum.
32276         (vec_perm_indices::series_p): Handle polynomial element counts.
32277         (vec_perm_indices::all_in_range_p): Likewise.
32278         (vec_perm_indices_to_tree): Likewise.
32279         (vec_perm_indices_to_rtx): Likewise.
32280         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
32281         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
32282         (tree_vector_builder::new_binary_operation): Handle polynomial
32283         element counts.  Return false if we need to know the number
32284         of elements at compile time.
32285         * fold-const.c (fold_vec_perm): Punt if the number of elements
32286         isn't known at compile time.
32288 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32290         * vec-perm-indices.h (vec_perm_builder): Change element type
32291         from HOST_WIDE_INT to poly_int64.
32292         (vec_perm_indices::element_type): Update accordingly.
32293         (vec_perm_indices::clamp): Handle polynomial element_types.
32294         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32295         (vec_perm_indices::all_in_range_p): Likewise.
32296         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
32297         than shwi trees.
32298         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
32299         polynomial vec_perm_indices element types.
32300         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
32301         * fold-const.c (fold_vec_perm): Likewise.
32302         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
32303         * tree-vect-generic.c (lower_vec_perm): Likewise.
32304         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32305         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
32306         element type to HOST_WIDE_INT.
32308 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32309             Alan Hayward  <alan.hayward@arm.com>
32310             David Sherwood  <david.sherwood@arm.com>
32312         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
32313         rather than an int.  Use plus_constant.
32314         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
32315         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
32317 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32318             Alan Hayward  <alan.hayward@arm.com>
32319             David Sherwood  <david.sherwood@arm.com>
32321         * calls.c (emit_call_1, expand_call): Change struct_value_size from
32322         a HOST_WIDE_INT to a poly_int64.
32324 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32325             Alan Hayward  <alan.hayward@arm.com>
32326             David Sherwood  <david.sherwood@arm.com>
32328         * calls.c (load_register_parameters): Cope with polynomial
32329         mode sizes.  Require a constant size for BLKmode parameters
32330         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
32331         forces a parameter to be padded at the lsb end in order to
32332         fill a complete number of words, require the parameter size
32333         to be ordered wrt UNITS_PER_WORD.
32335 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32336             Alan Hayward  <alan.hayward@arm.com>
32337             David Sherwood  <david.sherwood@arm.com>
32339         * reload1.c (spill_stack_slot_width): Change element type
32340         from unsigned int to poly_uint64_pod.
32341         (alter_reg): Treat mode sizes as polynomial.
32343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32344             Alan Hayward  <alan.hayward@arm.com>
32345             David Sherwood  <david.sherwood@arm.com>
32347         * reload.c (complex_word_subreg_p): New function.
32348         (reload_inner_reg_of_subreg, push_reload): Use it.
32350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32351             Alan Hayward  <alan.hayward@arm.com>
32352             David Sherwood  <david.sherwood@arm.com>
32354         * lra-constraints.c (process_alt_operands): Reject matched
32355         operands whose sizes aren't ordered.
32356         (match_reload): Refer to this check here.
32358 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32359             Alan Hayward  <alan.hayward@arm.com>
32360             David Sherwood  <david.sherwood@arm.com>
32362         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
32363         that the mode size is in the set {1, 2, 4, 8, 16}.
32365 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32366             Alan Hayward  <alan.hayward@arm.com>
32367             David Sherwood  <david.sherwood@arm.com>
32369         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
32370         Use plus_constant instead of gen_rtx_PLUS.
32372 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32373             Alan Hayward  <alan.hayward@arm.com>
32374             David Sherwood  <david.sherwood@arm.com>
32376         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
32377         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
32378         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
32379         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
32380         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
32381         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
32382         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
32383         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
32384         * config/i386/i386.c (ix86_push_rounding): ...this new function.
32385         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
32386         a poly_int64.
32387         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
32388         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
32389         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
32390         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
32391         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
32392         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
32393         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
32394         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
32395         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
32396         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
32397         function.
32398         * expr.c (emit_move_resolve_push): Treat the input and result
32399         of PUSH_ROUNDING as a poly_int64.
32400         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
32401         (emit_push_insn): Likewise.
32402         * lra-eliminations.c (mark_not_eliminable): Likewise.
32403         * recog.c (push_operand): Likewise.
32404         * reload1.c (elimination_effects): Likewise.
32405         * rtlanal.c (nonzero_bits1): Likewise.
32406         * calls.c (store_one_arg): Likewise.  Require the padding to be
32407         known at compile time.
32409 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32410             Alan Hayward  <alan.hayward@arm.com>
32411             David Sherwood  <david.sherwood@arm.com>
32413         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
32414         Use plus_constant instead of gen_rtx_PLUS.
32416 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32417             Alan Hayward  <alan.hayward@arm.com>
32418             David Sherwood  <david.sherwood@arm.com>
32420         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
32421         rather than an int.
32423 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32424             Alan Hayward  <alan.hayward@arm.com>
32425             David Sherwood  <david.sherwood@arm.com>
32427         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
32428         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
32429         via stack temporaries.  Treat the mode size as polynomial too.
32431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32432             Alan Hayward  <alan.hayward@arm.com>
32433             David Sherwood  <david.sherwood@arm.com>
32435         * expr.c (expand_expr_real_2): When handling conversions involving
32436         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
32437         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
32438         as a poly_uint64 too.
32440 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32441             Alan Hayward  <alan.hayward@arm.com>
32442             David Sherwood  <david.sherwood@arm.com>
32444         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
32446 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32447             Alan Hayward  <alan.hayward@arm.com>
32448             David Sherwood  <david.sherwood@arm.com>
32450         * combine.c (can_change_dest_mode): Handle polynomial
32451         REGMODE_NATURAL_SIZE.
32452         * expmed.c (store_bit_field_1): Likewise.
32453         * expr.c (store_constructor): Likewise.
32454         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
32455         and polynomial REGMODE_NATURAL_SIZE.
32456         (gen_lowpart_common): Likewise.
32457         * reginfo.c (record_subregs_of_mode): Likewise.
32458         * rtlanal.c (read_modify_subreg_p): Likewise.
32460 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32461             Alan Hayward  <alan.hayward@arm.com>
32462             David Sherwood  <david.sherwood@arm.com>
32464         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
32465         numbers of elements.
32467 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32468             Alan Hayward  <alan.hayward@arm.com>
32469             David Sherwood  <david.sherwood@arm.com>
32471         * match.pd: Cope with polynomial numbers of vector elements.
32473 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32474             Alan Hayward  <alan.hayward@arm.com>
32475             David Sherwood  <david.sherwood@arm.com>
32477         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
32478         in a POINTER_PLUS_EXPR.
32480 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32481             Alan Hayward  <alan.hayward@arm.com>
32482             David Sherwood  <david.sherwood@arm.com>
32484         * omp-simd-clone.c (simd_clone_subparts): New function.
32485         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
32486         (ipa_simd_modify_function_body): Likewise.
32488 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32489             Alan Hayward  <alan.hayward@arm.com>
32490             David Sherwood  <david.sherwood@arm.com>
32492         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
32493         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
32494         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
32495         (expand_vector_condition, vector_element): Likewise.
32496         (subparts_gt): New function.
32497         (get_compute_type): Use subparts_gt.
32498         (count_type_subparts): Delete.
32499         (expand_vector_operations_1): Use subparts_gt instead of
32500         count_type_subparts.
32502 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32503             Alan Hayward  <alan.hayward@arm.com>
32504             David Sherwood  <david.sherwood@arm.com>
32506         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
32507         (vect_compile_time_alias): ...this new function.  Do the calculation
32508         on poly_ints rather than trees.
32509         (vect_prune_runtime_alias_test_list): Update call accordingly.
32511 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32512             Alan Hayward  <alan.hayward@arm.com>
32513             David Sherwood  <david.sherwood@arm.com>
32515         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
32516         numbers of units.
32517         (vect_schedule_slp_instance): Likewise.
32519 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32520             Alan Hayward  <alan.hayward@arm.com>
32521             David Sherwood  <david.sherwood@arm.com>
32523         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
32524         constant and extern definitions for variable-length vectors.
32525         (vect_get_constant_vectors): Note that the number of units
32526         is known to be constant.
32528 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32529             Alan Hayward  <alan.hayward@arm.com>
32530             David Sherwood  <david.sherwood@arm.com>
32532         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
32533         of units as polynomial.  Choose between WIDE and NARROW based
32534         on multiple_p.
32536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32537             Alan Hayward  <alan.hayward@arm.com>
32538             David Sherwood  <david.sherwood@arm.com>
32540         * tree-vect-stmts.c (simd_clone_subparts): New function.
32541         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
32543 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32544             Alan Hayward  <alan.hayward@arm.com>
32545             David Sherwood  <david.sherwood@arm.com>
32547         * tree-vect-stmts.c (vectorizable_call): Treat the number of
32548         vectors as polynomial.  Use build_index_vector for
32549         IFN_GOMP_SIMD_LANE.
32551 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32552             Alan Hayward  <alan.hayward@arm.com>
32553             David Sherwood  <david.sherwood@arm.com>
32555         * tree-vect-stmts.c (get_load_store_type): Treat the number of
32556         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
32557         for variable-length vectors.
32558         (vectorizable_mask_load_store): Treat the number of units as
32559         polynomial, asserting that it is constant if the condition has
32560         already been enforced.
32561         (vectorizable_store, vectorizable_load): Likewise.
32563 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32564             Alan Hayward  <alan.hayward@arm.com>
32565             David Sherwood  <david.sherwood@arm.com>
32567         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
32568         of units as polynomial.  Punt if we can't tell at compile time
32569         which vector contains the final result.
32571 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32572             Alan Hayward  <alan.hayward@arm.com>
32573             David Sherwood  <david.sherwood@arm.com>
32575         * tree-vect-loop.c (vectorizable_induction): Treat the number
32576         of units as polynomial.  Punt on SLP inductions.  Use an integer
32577         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
32578         cast of such a series for variable-length floating-point
32579         reductions.
32581 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32582             Alan Hayward  <alan.hayward@arm.com>
32583             David Sherwood  <david.sherwood@arm.com>
32585         * tree.h (build_index_vector): Declare.
32586         * tree.c (build_index_vector): New function.
32587         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
32588         of units as polynomial, forcibly converting it to a constant if
32589         vectorizable_reduction has already enforced the condition.
32590         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
32591         to create a {1,2,3,...} vector.
32592         (vectorizable_reduction): Treat the number of units as polynomial.
32593         Choose vectype_in based on the largest scalar element size rather
32594         than the smallest number of units.  Enforce the restrictions
32595         relied on above.
32597 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32598             Alan Hayward  <alan.hayward@arm.com>
32599             David Sherwood  <david.sherwood@arm.com>
32601         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
32602         number of units as polynomial.
32604 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32605             Alan Hayward  <alan.hayward@arm.com>
32606             David Sherwood  <david.sherwood@arm.com>
32608         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
32609         * target.def (autovectorize_vector_sizes): Return the vector sizes
32610         by pointer, using vector_sizes rather than a bitmask.
32611         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
32612         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
32613         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
32614         Likewise.
32615         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
32616         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
32617         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
32618         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
32619         * omp-general.c (omp_max_vf): Likewise.
32620         * omp-low.c (omp_clause_aligned_alignment): Likewise.
32621         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
32622         * tree-vect-loop.c (vect_analyze_loop): Likewise.
32623         * tree-vect-slp.c (vect_slp_bb): Likewise.
32624         * doc/tm.texi: Regenerate.
32625         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
32626         to a poly_uint64.
32627         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
32628         the vector size as a poly_uint64 rather than an unsigned int.
32629         (current_vector_size): Change from an unsigned int to a poly_uint64.
32630         (get_vectype_for_scalar_type): Update accordingly.
32631         * tree.h (build_truth_vector_type): Take the size and number of
32632         units as a poly_uint64 rather than an unsigned int.
32633         (build_vector_type): Add a temporary overload that takes
32634         the number of units as a poly_uint64 rather than an unsigned int.
32635         * tree.c (make_vector_type): Likewise.
32636         (build_truth_vector_type): Take the number of units as a poly_uint64
32637         rather than an unsigned int.
32639 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32640             Alan Hayward  <alan.hayward@arm.com>
32641             David Sherwood  <david.sherwood@arm.com>
32643         * target.def (get_mask_mode): Take the number of units and length
32644         as poly_uint64s rather than unsigned ints.
32645         * targhooks.h (default_get_mask_mode): Update accordingly.
32646         * targhooks.c (default_get_mask_mode): Likewise.
32647         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
32648         * doc/tm.texi: Regenerate.
32650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32651             Alan Hayward  <alan.hayward@arm.com>
32652             David Sherwood  <david.sherwood@arm.com>
32654         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
32655         * omp-general.c (omp_max_vf): Likewise.
32656         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
32657         (expand_omp_simd): Handle polynomial safelen.
32658         * omp-low.c (omplow_simd_context): Add a default constructor.
32659         (omplow_simd_context::max_vf): Change from int to poly_uint64.
32660         (lower_rec_simd_input_clauses): Update accordingly.
32661         (lower_rec_input_clauses): Likewise.
32663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32664             Alan Hayward  <alan.hayward@arm.com>
32665             David Sherwood  <david.sherwood@arm.com>
32667         * tree-vectorizer.h (vect_nunits_for_cost): New function.
32668         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
32669         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
32670         (vect_analyze_slp_cost): Likewise.
32671         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
32672         (vect_model_load_cost): Likewise.
32674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32675             Alan Hayward  <alan.hayward@arm.com>
32676             David Sherwood  <david.sherwood@arm.com>
32678         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
32679         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
32680         from an unsigned int * to a poly_uint64_pod *.
32681         (calculate_unrolling_factor): New function.
32682         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
32684 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32685             Alan Hayward  <alan.hayward@arm.com>
32686             David Sherwood  <david.sherwood@arm.com>
32688         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32689         from an unsigned int to a poly_uint64.
32690         (_loop_vec_info::slp_unrolling_factor): Likewise.
32691         (_loop_vec_info::vectorization_factor): Change from an int
32692         to a poly_uint64.
32693         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32694         (vect_get_num_vectors): New function.
32695         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32696         (vect_get_num_copies): Use vect_get_num_vectors.
32697         (vect_analyze_data_ref_dependences): Change max_vf from an int *
32698         to an unsigned int *.
32699         (vect_analyze_data_refs): Change min_vf from an int * to a
32700         poly_uint64 *.
32701         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32702         than an unsigned HOST_WIDE_INT.
32703         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32704         (vect_analyze_data_ref_dependence): Change max_vf from an int *
32705         to an unsigned int *.
32706         (vect_analyze_data_ref_dependences): Likewise.
32707         (vect_compute_data_ref_alignment): Handle polynomial vf.
32708         (vect_enhance_data_refs_alignment): Likewise.
32709         (vect_prune_runtime_alias_test_list): Likewise.
32710         (vect_shift_permute_load_chain): Likewise.
32711         (vect_supportable_dr_alignment): Likewise.
32712         (dependence_distance_ge_vf): Take the vectorization factor as a
32713         poly_uint64 rather than an unsigned HOST_WIDE_INT.
32714         (vect_analyze_data_refs): Change min_vf from an int * to a
32715         poly_uint64 *.
32716         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32717         vfm1 as a poly_uint64 rather than an int.  Make the same change
32718         for the returned bound_scalar.
32719         (vect_gen_vector_loop_niters): Handle polynomial vf.
32720         (vect_do_peeling): Likewise.  Update call to
32721         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32722         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32723         be constant.
32724         * tree-vect-loop.c (vect_determine_vectorization_factor)
32725         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32726         (vect_get_known_peeling_cost): Likewise.
32727         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32728         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32729         (vect_transform_loop): Likewise.  Use the lowest possible VF when
32730         updating the upper bounds of the loop.
32731         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
32732         rather than an int.
32733         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32734         polynomial unroll factors.
32735         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32736         (vect_make_slp_decision): Likewise.
32737         (vect_supported_load_permutation_p): Likewise, and polynomial
32738         vf too.
32739         (vect_analyze_slp_cost): Handle polynomial vf.
32740         (vect_slp_analyze_node_operations): Likewise.
32741         (vect_slp_analyze_bb_1): Likewise.
32742         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32743         than an unsigned HOST_WIDE_INT.
32744         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32745         (vectorizable_load): Handle polynomial vf.
32746         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32747         a poly_uint64.
32748         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32750 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32751             Alan Hayward  <alan.hayward@arm.com>
32752             David Sherwood  <david.sherwood@arm.com>
32754         * match.pd: Handle bit operations involving three constants
32755         and try to fold one pair.
32757 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32759         * tree-vect-loop-manip.c: Include gimple-fold.h.
32760         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32761         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
32762         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32763         Add a path that uses a step of VF instead of 1, but disable it
32764         for now.
32765         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32766         and niters_no_overflow parameters.  Update calls to
32767         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32768         Create a new SSA name if the latter choses to use a ste other
32769         than zero, and return it via niters_vector_mult_vf_var.
32770         * tree-vect-loop.c (vect_transform_loop): Update calls to
32771         vect_do_peeling, vect_gen_vector_loop_niters and
32772         slpeel_make_loop_iterate_ntimes.
32773         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32774         (vect_gen_vector_loop_niters): Update declarations after above changes.
32776 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
32778         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32779         128-bit round to integer instructions.
32780         (ceil<mode>2): Likewise.
32781         (btrunc<mode>2): Likewise.
32782         (round<mode>2): Likewise.
32784 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32786         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32787         unaligned VSX load/store on P8/P9.
32788         (expand_block_clear): Allow the use of unaligned VSX
32789         load/store on P8/P9.
32791 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32793         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32794         New function.
32795         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32796         swap associated with both a load and a store.
32798 2018-01-02  Andrew Waterman  <andrew@sifive.com>
32800         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32801         * config/riscv/riscv.md (clear_cache): Use it.
32803 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
32805         * web.c: Remove out-of-date comment.
32807 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32809         * expr.c (fixup_args_size_notes): Check that any existing
32810         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32811         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32812         (emit_single_push_insn): ...here.
32814 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32816         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32817         (const_vector_encoded_nelts): New function.
32818         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32819         (const_vector_int_elt, const_vector_elt): Declare.
32820         * emit-rtl.c (const_vector_int_elt_1): New function.
32821         (const_vector_elt): Likewise.
32822         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32823         of CONST_VECTOR_ELT.
32825 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32827         * expr.c: Include rtx-vector-builder.h.
32828         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32829         directly on the tree encoding.
32830         (const_vector_from_tree): Likewise.
32831         * optabs.c: Include rtx-vector-builder.h.
32832         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32833         sequence of "u" values.
32834         * vec-perm-indices.c: Include rtx-vector-builder.h.
32835         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32836         directly on the vec_perm_indices encoding.
32838 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32840         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32841         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32842         * rtx-vector-builder.h: New file.
32843         * rtx-vector-builder.c: Likewise.
32844         * rtl.h (rtx_def::u2): Add a const_vector field.
32845         (CONST_VECTOR_NPATTERNS): New macro.
32846         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32847         (CONST_VECTOR_DUPLICATE_P): Likewise.
32848         (CONST_VECTOR_STEPPED_P): Likewise.
32849         (CONST_VECTOR_ENCODED_ELT): Likewise.
32850         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32851         (unwrap_const_vec_duplicate): Likewise.
32852         (const_vec_series_p): Check for a non-duplicated vector encoding.
32853         Say that the function only returns true for integer vectors.
32854         * emit-rtl.c: Include rtx-vector-builder.h.
32855         (gen_const_vec_duplicate_1): Delete.
32856         (gen_const_vector): Call gen_const_vec_duplicate instead of
32857         gen_const_vec_duplicate_1.
32858         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32859         (gen_const_vec_duplicate): Use rtx_vector_builder.
32860         (gen_const_vec_series): Likewise.
32861         (gen_rtx_CONST_VECTOR): Likewise.
32862         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32863         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32864         Build a new vector rather than modifying a CONST_VECTOR in-place.
32865         (handle_special_swappables): Update call accordingly.
32866         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32867         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32868         Build a new vector rather than modifying a CONST_VECTOR in-place.
32869         (handle_special_swappables): Update call accordingly.
32871 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32873         * simplify-rtx.c (simplify_const_binary_operation): Use
32874         CONST_VECTOR_ELT instead of XVECEXP.
32876 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32878         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32879         the selector elements to be different from the data elements
32880         if the selector is a VECTOR_CST.
32881         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32882         ssizetype for the selector.
32884 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32886         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32887         before testing each element individually.
32888         * tree-vect-generic.c (lower_vec_perm): Likewise.
32890 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32892         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32893         * selftest-run-tests.c (selftest::run_tests): Call it.
32894         * vector-builder.h (vector_builder::operator ==): New function.
32895         (vector_builder::operator !=): Likewise.
32896         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32897         (vec_perm_indices::all_from_input_p): New function.
32898         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32899         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32900         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32901         instead of reading the VECTOR_CST directly.  Detect whether both
32902         vector inputs are the same before constructing the vec_perm_indices,
32903         and update the number of inputs argument accordingly.  Use the
32904         utility functions added above.  Only construct sel2 if we need to.
32906 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32908         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32909         the broadcast of the low byte.
32910         (expand_mult_highpart): Use an explicit encoding for the permutes.
32911         * optabs-query.c (can_mult_highpart_p): Likewise.
32912         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32913         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32914         (vectorizable_bswap): Likewise.
32915         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32916         explicit encoding for the power-of-2 permutes.
32917         (vect_permute_store_chain): Likewise.
32918         (vect_grouped_load_supported): Likewise.
32919         (vect_permute_load_chain): Likewise.
32921 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32923         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32924         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32925         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32926         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32927         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32928         (vect_gen_perm_mask_any): Likewise.
32930 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32932         * int-vector-builder.h: New file.
32933         * vec-perm-indices.h: Include int-vector-builder.h.
32934         (vec_perm_indices): Redefine as an int_vector_builder.
32935         (auto_vec_perm_indices): Delete.
32936         (vec_perm_builder): Redefine as a stand-alone class.
32937         (vec_perm_indices::vec_perm_indices): New function.
32938         (vec_perm_indices::clamp): Likewise.
32939         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32940         (vec_perm_indices::new_vector): New function.
32941         (vec_perm_indices::new_expanded_vector): Update for new
32942         vec_perm_indices class.
32943         (vec_perm_indices::rotate_inputs): New function.
32944         (vec_perm_indices::all_in_range_p): Operate directly on the
32945         encoded form, without computing elided elements.
32946         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32947         encoding.  Update for new vec_perm_indices class.
32948         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32949         the given vec_perm_builder.
32950         (expand_vec_perm_var): Update vec_perm_builder constructor.
32951         (expand_mult_highpart): Use vec_perm_builder instead of
32952         auto_vec_perm_indices.
32953         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32954         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32955         or double series encoding as appropriate.
32956         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32957         vec_perm_indices instead of auto_vec_perm_indices.
32958         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32959         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32960         (vect_permute_store_chain): Likewise.
32961         (vect_grouped_load_supported): Likewise.
32962         (vect_permute_load_chain): Likewise.
32963         (vect_shift_permute_load_chain): Likewise.
32964         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32965         (vect_transform_slp_perm_load): Likewise.
32966         (vect_schedule_slp_instance): Likewise.
32967         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32968         (vectorizable_mask_load_store): Likewise.
32969         (vectorizable_bswap): Likewise.
32970         (vectorizable_store): Likewise.
32971         (vectorizable_load): Likewise.
32972         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32973         vec_perm_indices instead of auto_vec_perm_indices.  Use
32974         tree_to_vec_perm_builder to read the vector from a tree.
32975         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32976         vec_perm_builder instead of a vec_perm_indices.
32977         (have_whole_vector_shift): Use vec_perm_builder and
32978         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32979         truncation to calc_vec_perm_mask_for_shift.
32980         (vect_create_epilog_for_reduction): Likewise.
32981         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32982         from auto_vec_perm_indices to vec_perm_indices.
32983         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32984         instead of changing individual elements.
32985         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32986         the vector in d.perm.
32987         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32988         from auto_vec_perm_indices to vec_perm_indices.
32989         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32990         instead of changing individual elements.
32991         (arm_vectorize_vec_perm_const): Use new_vector to install
32992         the vector in d.perm.
32993         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32994         Update vec_perm_builder constructor.
32995         (rs6000_expand_interleave): Likewise.
32996         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32997         (rs6000_expand_interleave): Likewise.
32999 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33001         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
33002         to qimode could truncate the indices.
33003         * optabs.c (expand_vec_perm_var): Likewise.
33005 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33007         * Makefile.in (OBJS): Add vec-perm-indices.o.
33008         * vec-perm-indices.h: New file.
33009         * vec-perm-indices.c: Likewise.
33010         * target.h (vec_perm_indices): Replace with a forward class
33011         declaration.
33012         (auto_vec_perm_indices): Move to vec-perm-indices.h.
33013         * optabs.h: Include vec-perm-indices.h.
33014         (expand_vec_perm): Delete.
33015         (selector_fits_mode_p, expand_vec_perm_var): Declare.
33016         (expand_vec_perm_const): Declare.
33017         * target.def (vec_perm_const_ok): Replace with...
33018         (vec_perm_const): ...this new hook.
33019         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
33020         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
33021         * doc/tm.texi: Regenerate.
33022         * optabs.def (vec_perm_const): Delete.
33023         * doc/md.texi (vec_perm_const): Likewise.
33024         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
33025         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
33026         expand_vec_perm for constant permutation vectors.  Assert that
33027         the mode of variable permutation vectors is the integer equivalent
33028         of the mode that is being permuted.
33029         * optabs-query.h (selector_fits_mode_p): Declare.
33030         * optabs-query.c: Include vec-perm-indices.h.
33031         (selector_fits_mode_p): New function.
33032         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
33033         is defined, instead of checking whether the vec_perm_const_optab
33034         exists.  Use targetm.vectorize.vec_perm_const instead of
33035         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
33036         fit in the vector mode before using a variable permute.
33037         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
33038         vec_perm_indices instead of an rtx.
33039         (expand_vec_perm): Replace with...
33040         (expand_vec_perm_const): ...this new function.  Take the selector
33041         as a vec_perm_indices rather than an rtx.  Also take the mode of
33042         the selector.  Update call to shift_amt_for_vec_perm_mask.
33043         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
33044         Use vec_perm_indices::new_expanded_vector to expand the original
33045         selector into bytes.  Check whether the indices fit in the vector
33046         mode before using a variable permute.
33047         (expand_vec_perm_var): Make global.
33048         (expand_mult_highpart): Use expand_vec_perm_const.
33049         * fold-const.c: Includes vec-perm-indices.h.
33050         * tree-ssa-forwprop.c: Likewise.
33051         * tree-vect-data-refs.c: Likewise.
33052         * tree-vect-generic.c: Likewise.
33053         * tree-vect-loop.c: Likewise.
33054         * tree-vect-slp.c: Likewise.
33055         * tree-vect-stmts.c: Likewise.
33056         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
33057         Delete.
33058         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
33059         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
33060         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
33061         (aarch64_vectorize_vec_perm_const): ...this new function.
33062         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33063         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33064         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
33065         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
33066         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33067         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33068         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
33069         into...
33070         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
33071         check for NEON modes.
33072         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
33073         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
33074         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
33075         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
33076         into...
33077         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
33078         the old VEC_PERM_CONST conditions.
33079         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
33080         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
33081         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
33082         (ia64_vectorize_vec_perm_const_ok): Merge into...
33083         (ia64_vectorize_vec_perm_const): ...this new function.
33084         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
33085         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
33086         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
33087         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
33088         * config/mips/mips.c (mips_expand_vec_perm_const)
33089         (mips_vectorize_vec_perm_const_ok): Merge into...
33090         (mips_vectorize_vec_perm_const): ...this new function.
33091         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
33092         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
33093         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
33094         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
33095         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
33096         (rs6000_expand_vec_perm_const): Delete.
33097         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
33098         Delete.
33099         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33100         (altivec_expand_vec_perm_const_le): Take each operand individually.
33101         Operate on constant selectors rather than rtxes.
33102         (altivec_expand_vec_perm_const): Likewise.  Update call to
33103         altivec_expand_vec_perm_const_le.
33104         (rs6000_expand_vec_perm_const): Delete.
33105         (rs6000_vectorize_vec_perm_const_ok): Delete.
33106         (rs6000_vectorize_vec_perm_const): New function.
33107         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33108         an element count and rtx array.
33109         (rs6000_expand_extract_even): Update call accordingly.
33110         (rs6000_expand_interleave): Likewise.
33111         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
33112         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
33113         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
33114         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
33115         (rs6000_expand_vec_perm_const): Delete.
33116         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33117         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33118         (altivec_expand_vec_perm_const_le): Take each operand individually.
33119         Operate on constant selectors rather than rtxes.
33120         (altivec_expand_vec_perm_const): Likewise.  Update call to
33121         altivec_expand_vec_perm_const_le.
33122         (rs6000_expand_vec_perm_const): Delete.
33123         (rs6000_vectorize_vec_perm_const_ok): Delete.
33124         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
33125         reference to the SPE evmerge intructions.
33126         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33127         an element count and rtx array.
33128         (rs6000_expand_extract_even): Update call accordingly.
33129         (rs6000_expand_interleave): Likewise.
33130         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
33131         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
33132         new function.
33133         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33135 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33137         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
33138         vector mode and that that mode matches the mode of the data
33139         being permuted.
33140         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
33141         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
33142         directly using expand_vec_perm_1 when forcing selectors into
33143         registers.
33144         (expand_vec_perm_var): New function, split out from expand_vec_perm.
33146 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33148         * optabs-query.h (can_vec_perm_p): Delete.
33149         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
33150         * optabs-query.c (can_vec_perm_p): Split into...
33151         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
33152         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
33153         particular selector is valid.
33154         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33155         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33156         (vect_grouped_load_supported): Likewise.
33157         (vect_shift_permute_load_chain): Likewise.
33158         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
33159         (vect_transform_slp_perm_load): Likewise.
33160         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33161         (vectorizable_bswap): Likewise.
33162         (vect_gen_perm_mask_checked): Likewise.
33163         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
33164         implementations of variable permutation vectors into account
33165         when deciding which selector to use.
33166         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
33167         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
33168         with a false third argument.
33169         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
33170         to test whether the constant selector is valid and can_vec_perm_var_p
33171         to test whether a variable selector is valid.
33173 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33175         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
33176         * optabs-query.c (can_vec_perm_p): Likewise.
33177         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
33178         instead of vec_perm_indices.
33179         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
33180         (vect_gen_perm_mask_checked): Likewise,
33181         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
33182         (vect_gen_perm_mask_checked): Likewise,
33184 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33186         * optabs-query.h (qimode_for_vec_perm): Declare.
33187         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
33188         (qimode_for_vec_perm): ...this new function.
33189         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
33191 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
33193         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
33194         does not have a conditional at the top.
33196 2018-01-02  Richard Biener  <rguenther@suse.de>
33198         * ipa-inline.c (big_speedup_p): Fix expression.
33200 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
33202         PR target/81616
33203         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
33204         for generic 4->6.
33206 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
33208         PR target/81616
33209         Generic tuning.
33210         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
33211         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
33212         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
33213         cond_taken_branch_cost 3->4.
33215 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
33217         PR tree-optimization/83581
33218         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
33219         TODO_cleanup_cfg if any changes have been made.
33221         PR middle-end/83608
33222         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
33223         convert_modes if target mode has the right side, but different mode
33224         class.
33226         PR middle-end/83609
33227         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
33228         last argument when extracting from CONCAT.  If either from_real or
33229         from_imag is NULL, use expansion through memory.  If result is not
33230         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
33231         the parts directly to inner mode, if even that fails, use expansion
33232         through memory.
33234         PR middle-end/83623
33235         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
33236         check for bswap in mode rather than HImode and use that in expand_unop
33237         too.
33239 Copyright (C) 2018 Free Software Foundation, Inc.
33241 Copying and distribution of this file, with or without modification,
33242 are permitted in any medium without royalty provided the copyright
33243 notice and this notice are preserved.