Fix ICE in lto_symtab_merge_symbols_1 (PR lto/88004).
[official-gcc.git] / gcc / ChangeLog
blob3f7e443e57da94e14867a8ba57cabb2f052fc682
1 2018-11-16  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/88011
4         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
5         replacing set_value_range_to_undefined and
6         set_value_range_to_varying with method calls.
8 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10         * config/s390/s390.md
11         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
12         assembler.
13         (*r<noxa>sbg_<mode>_sll): Likewise.
14         (*r<noxa>sbg_<mode>_srl): Likewise.
15         (*r<noxa>sbg_sidi_srl): New pattern.
17 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
19         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
20         Default to TARGET_VXWORKS_RTP.
21         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
22         of TARGET_VXWORKS_RTP.
23         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
24         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
25         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
27 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
29         PR c++/25759
30         * doc/extend.texi (Common Type Attributes): Make it explicit
31         that attribute "packed" can apply to C++ classes.
33 2018-11-15  Martin Sebor  <msebor@redhat.com>
35         PR c++/87541
36         PR c++/87542
37         * tree.c (type_argument_type): New function.
38         * tree.h (type_argument_type): Declare it.
39         * gcc/doc/extend.texi (alloc_align): Update and clarify.
40         (alloc_size, nonnull, sentinel): Same.
42 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
43             Kwok Cheung Yeung  <kcy@codesourcery.com>
45         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
46         int_mode_for_size fails.
47         (vectorizable_load): Likewise.
49 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
51         * doc/ux.texi (Group logically-related diagnostics): Move
52         discussion of auto_diagnostic_group into this new subsection.
53         Give an example of where this grouping is used.
55 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
57         PR other/19165
58         * Makefile.in (OBJS): Move json.o to...
59         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
60         * common.opt (fdiagnostics-format=): New option.
61         (diagnostics_output_format): New enum.
62         * diagnostic-format-json.cc: New file.
63         * diagnostic.c (default_diagnostic_final_cb): New function, taken
64         from start of diagnostic_finish.
65         (diagnostic_initialize): Initialize final_cb to
66         default_diagnostic_final_cb.
67         (diagnostic_finish): Move "being treated as errors" messages to
68         default_diagnostic_final_cb.  Call any final_cb.
69         (default_diagnostic_finalizer): Add diagnostic_t param.
70         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
71         diagnostic_finalizer callback.
72         * diagnostic.h (enum diagnostics_output_format): New enum.
73         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
74         (struct diagnostic_context): Add "final_cb".
75         (default_diagnostic_finalizer): Add diagnostic_t param.
76         (diagnostic_output_format_init): New decl.
77         * doc/invoke.texi (-fdiagnostics-format): New option.
78         * dwarf2out.c (gen_producer_string): Ignore
79         OPT_fdiagnostics_format_.
80         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
81         * lto-wrapper.c (append_diag_options): Ignore it.
82         * opts.c (common_handle_option): Handle it.
84 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
86         PR tree-optimization/88015
87         * graphite-isl-ast-to-gimple.c
88         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
89         for dump_enabled_p.
90         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
92 2018-11-15  Richard Biener  <rguenther@suse.de>
94         PR middle-end/88029
95         * gimple.c (gimple_call_flags): Union flags from decl, type
96         and call fntype.
97         * trans-mem.c (is_tm_pure_call): Simplify.
99 2018-11-15  Richard Biener  <rguenther@suse.de>
101         PR tree-optimization/88031
102         * tree-vect-loop.c (vectorizable_reduction): Move check
103         for multiple types earlier so we get the expected dump.
104         Simplify calls to vectorizable_condition.
105         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
106         * tree-vect-stmts.c (vectorizable_condition): Instead of
107         reduc_def and reduc_index take just a flag.  Simplify
108         code-generation now that we can rely on the defs being set up.
109         (vectorizable_comparison): Remove unused argument.
111 2018-11-15  Richard Biener  <rguenther@suse.de>
113         PR tree-optimization/88030
114         * tree-complex.c (need_eh_cleanup): New global.
115         (update_complex_assignment): Mark blocks that need EH update.
116         (expand_complex_comparison): Likewise.
117         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
118         perform EH cleanup and schedule CFG cleanup if that did anything.
120 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
122         PR rtl-optimization/88018
123         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
124         on the presence of fallthru edge, rather than if it is a USE or not.
126 2018-11-15  Richard Biener  <rguenther@suse.de>
128         PR middle-end/87917
129         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
130         analyze_subscript_affine_affine properly.
132 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
134         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
135         (GLIBC_DYNAMIC_LINKER): Define.
136         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
138 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
140         PR tree-optimization/84648
141         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
142         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
143         by calling adjust_cond_for_loop_until_wrap.
145 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
147         PR other/56334
148         * doc/extend.texi (Common Function Attributes): Clarify linker
149         restrictions on "aligned" attribute.
150         (Common Variable Attributes): Likewise.  Mention that linker
151         restrictions don't apply to stack-allocated variables.
153 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
155         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
156         as an lvalue.
158 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
160         PR lto/55102
161         PR lto/56700
162         * doc/invoke.texi (Optimize Options): Remove bad example about
163         interaction between -flto and -O.  Replace it with a note that
164         you need to compile with -O and not just link.  Copy-edit -flto
165         discussion to reduce verbiage and improve flow.
167 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
169         PR middle-end/59658
170         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
171         also suppress many optimizations.  Alphabetize option lists for
172         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
173         correct documentation for those options to say that.
174         * opts.c (default_options_table): Sort table by level and option
175         name, to make it easier to correlate to the manual.
177 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
179         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
180         printing of the names of segment registers.
182 2018-11-14  Nathan Sidwell  <nathan@acm.org>
184         PR debug/88006
185         PR debug/87462
186         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
187         type list.
189 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
191         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
193 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
195         P1236R1 - Signed integers are two's complement
196         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
198         PR bootstrap/86739
199         * hash-map.h (hash_map::iterator::reference_pair): New class.
200         (hash_map::iterator::operator*): Return it rather than std::pair.
202 2018-11-14  Jeff Law  <law@redhat.com>
204         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
205         iff the operand is a constant.
207 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
209         * gimple-ssa-evrp-analyze.c
210         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
211         ignore_equivs_equal_p to equal_p.
212         * ipa-cp.c (meet_with_1): Use equal_p instead of
213         ignore_equivs_equal_p.
214         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
215         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
216         (value_range::operator==): Remove.
217         (value_range::operator!=): Remove.
218         (vrp_prop::visit_stmt): Use equal_p.
219         * tree-vrp.h (value_range): Remove operator==, operator!=,
220         ignore_equivs_equal_p.
221         * vr-values.c (update_value_range): Use equal_p.
223 2018-11-14  Michael Matz  <matz@suse.de>
225         PR middle-end/86575
226         * gimplify.c (collect_fallthrough_labels): Add new argument,
227         return location via that, don't modify statements.
228         (warn_implicit_fallthrough_r): Adjust call, don't use
229         statement location directly.
231 2018-11-14  Richard Biener  <rguenther@suse.de>
233         PR middle-end/87985
234         * tree-data-ref.c (split_constant_offset): Add wrapper
235         allocating a cache hash-map.
236         (split_constant_offset_1): Cache results of expanding
237         expressions from SSA def stmts.
239 2018-11-14  Richard Biener  <rguenther@suse.de>
241         PR middle-end/88021
242         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
243         to lambda_int.
244         (lambda_vector_mult_const): Likewise.
245         (lambda_matrix_right_hermite): Use lambda_int temporaries.
247 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
248             Jackson Woodruff  <jackson.woodruff@arm.com>
250         PR 71026/tree-optimization
251         * match.pd: Simplify floating point comparisons.
253 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
255         PR rtl-optimization/87817
256         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
257         last argument has low 8 bits clear, fold to 0.
259 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
261         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
262         MAX_OFILE_ALIGNMENT as needed.
264 2018-11-14  Richard Biener  <rguenther@suse.de>
266         PR tree-optimization/88019
267         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
268         COND_EXPR arguments to non-trapping overflow.
270 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
272         * config/arc/arc.h (reg_class): Reorder registers classes, remove
273         unused register classes.
274         (REG_CLASS_NAMES): Likewise.
275         (REG_CLASS_CONTENTS): Likewise.
276         (FIXED_REGISTERS): Make lp_count fixed.
277         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
278         (PROGRAM_COUNTER_REGNO): Remove.
279         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
280         register classes, use constants for register numbers, remove
281         reg_class_contents references.
282         (arc_process_double_reg_moves): Add asserts.
283         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
284         lra_in_progress predicate.
285         (arc_init_reg_tables): Remove unused register classes.
286         (arc_register_move_cost): Likewise.
287         (arc_preferred_reload_class): Likewise.
288         (hwloop_optimize): Update rtx patterns involving lp_count register.
289         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
290         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
291         CHEAP_CORE_REGS.  Former one will be used for LRA.
292         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
293         one will be used for LRA.
294         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
295         one will be used for LRA.
296         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
297         Former one will be used for LRA.
298         ("f"): Delete constraint.
299         ("k"): Likewise.
300         ("e"): Likewise.
301         ("l"): Change it from register constraint to constraint.
302         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
303         (movhi_insn): Likewise.
304         (movsi_insn): Update pattern.
305         (arc_lp): Likewise.
306         (dbnz): Likewise.
307         (stack_tie): Remove 'b' constraint letter.
308         (R4_REG): Define.
309         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
310         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
311         (R57_REG, R59_REG, PCL_REG): Likewise.
312         (ILINK1_REGNUM): Renamed to ILINK1_REG.
313         (ILINK2_REGNUM): Renamed to ILINK2_REG.
314         (Rgp): Remove.
315         (SP_REGS): Likewise.
316         (Rcw): Remove unused reg classes.
317         * config/arc/predicates.md (dest_reg_operand): Just default on
318         register_operand predicate.
319         (mpy_dest_reg_operand): Likewise.
320         (move_dest_operand): Use macros instead of constants.
321         * config/arc/arc.opt (mlra): Switch to lra as default.
323 2018-11-14  Richard Biener  <rguenther@suse.de>
325         PR tree-optimization/87974
326         * tree-vect-loop.c (vectorizable_reduction): When computing
327         the vectorized reduction PHI vector type ignore constant
328         and external defs.
330 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
332         PR tree-optimization/87977
333         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
334         stmt, build a new one and replace the old one with it.  Formatting fix.
335         Call release_ssa_name (x) if !has_other_use and !delete_div.
336         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
337         verify lhs of stmt is still def.
339 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
341         PR rtl-optimization/87507
342         * lower-subreg.c (operand_for_swap_move_operator): New function.
343         (simple_move): Strip simple operators.
344         (find_pseudo_copy): Likewise.
345         (resolve_operand_for_swap_move_operator): New function.
346         (resolve_simple_move): Strip simple operators and swap operands.
348 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
350         PR tree-optimization/87898
351         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
352         (ipa_simd_modify_function_body): Remove debug stmts where the first
353         argument was changed into a non-decl.
355 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
357         PR rtl-optimization/87899
358         * lra-lives.c (start_living): Update white space in comment.
359         (enum point_type): New.
360         (sparseset_contains_pseudos_p): New function.
361         (update_pseudo_point): Likewise.
362         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
363         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
364         handling.  Move early exit after adding conflicts.
365         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
366         if regno is already live.  Remove all handling of program points.
367         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
368         after adding conflicts.  Remove all handling of program points and
369         ignore_reg_for_conflicts.
370         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
371         and do not guard call to mark_pseudo_live.
372         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
373         and do not guard call to mark_pseudo_dead.
374         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
375         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
376         Use new function update_pseudo_point.  Handle register copies by
377         removing the source register from the live set.  Handle INOUT operands.
378         Update to the next program point using the unused_set, dead_set and
379         start_dying sets.
380         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
382 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
384         * builtins.c: Replace "source_location" with "location_t".
385         * diagnostic-show-locus.c: Likewise.
386         * diagnostic.c: Likewise.
387         * dumpfile.c: Likewise.
388         * gcc-rich-location.h: Likewise.
389         * genmatch.c: Likewise.
390         * gimple.h: Likewise.
391         * gimplify.c: Likewise.
392         * input.c: Likewise.
393         * input.h: Likewise.  Eliminate the typedef.
394         * omp-expand.c: Likewise.
395         * selftest.h: Likewise.
396         * substring-locations.h (get_source_location_for_substring):
397         Rename to...
398         (get_location_within_string): ...this.
399         * tree-cfg.c: Replace "source_location" with "location_t".
400         * tree-cfgcleanup.c: Likewise.
401         * tree-diagnostic.c: Likewise.
402         * tree-into-ssa.c: Likewise.
403         * tree-outof-ssa.c: Likewise.
404         * tree-parloops.c: Likewise.
405         * tree-phinodes.c: Likewise.
406         * tree-phinodes.h: Likewise.
407         * tree-ssa-loop-ivopts.c: Likewise.
408         * tree-ssa-loop-manip.c: Likewise.
409         * tree-ssa-phiopt.c: Likewise.
410         * tree-ssa-phiprop.c: Likewise.
411         * tree-ssa-threadupdate.c: Likewise.
412         * tree-ssa.c: Likewise.
413         * tree-ssa.h: Likewise.
414         * tree-vect-loop-manip.c: Likewise.
416 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
418         * doc/extend.texi: Fix typo in the weakref description.
420 2018-11-13  Richard Biener  <rguenther@suse.de>
422         PR tree-optimization/86991
423         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
424         group building until we have successfully detected the SLP
425         reduction.
426         (vect_is_simple_reduction): Remove fixup code here.
428 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
430         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
431         (dump_gimple_stmt): Use it.
432         (dump_gimple_stmt_loc): Likewise.
433         (dump_gimple_expr): Likewise.
434         (dump_gimple_expr_loc): Likewise.
435         (dump_generic_expr): Likewise.
436         (dump_generic_expr_loc): Likewise.
437         (dump_printf): Likewise.
438         (dump_printf_loc): Likewise.
439         (dump_dec): Likewise.
440         (dump_dec): Likewise.
441         (dump_hex): Likewise.
442         (dump_symtab_node): Likewise.
443         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
444         Guard dump call with dump_enabled_p.
445         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
446         * graphite-optimize-isl.c (optimize_isl): Likewise.
447         * graphite.c (graphite_transform_loops): Likewise.
448         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
449         * tree-parloops.c (parallelize_loops): Likewise.
450         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
451         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
452         (vect_prune_runtime_alias_test_list): Likewise.
453         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
454         (vect_estimate_min_profitable_iters): Likewise.
455         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
456         (vect_build_slp_tree_2): Likewise.
457         (vect_supported_load_permutation_p): Likewise.
458         (vect_slp_analyze_operations): Likewise.
459         (vect_slp_analyze_bb_1): Likewise.
460         (vect_slp_bb): Likewise.
461         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
462         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
463         (pass_slp_vectorize::execute): Likewise.
464         (increase_alignment): Likewise.
466 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
468         PR ipa/87955
469         * ipa-inline.c (report_inline_failed_reason): Guard calls to
470         cl_target_option_print_diff and cl_optimization_print_diff with
471         if (dump_file).
473 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
475         * doc/invoke.texi (-fsave-optimization-record): Note that the
476         output is compressed.
477         * optinfo-emit-json.cc: Include <zlib.h>.
478         (optrecord_json_writer::write): Compress the output.
480 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
482         * tree-vrp.c (value_range_base::dump): Dump type.
483         Do not use INF nomenclature for 1-bit types.
484         (dump_value_range): Group all variants to common dumping code.
485         (debug): New overloaded functions for value_ranges.
486         (value_range_base::dump): Remove no argument version.
487         (value_range::dump): Same.
489 2018-11-13  Richard Biener  <rguenther@suse.de>
491         PR tree-optimization/87931
492         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
493         nested cycles we support to latch computations vectorizable_reduction
494         handles.
496 2018-11-13  Martin Liska  <mliska@suse.cz>
498         PR tree-optimization/87885
499         * cfghooks.c (account_profile_record): Rename
500         to ...
501         (profile_record_check_consistency): ... this.
502         Calculate missing num_mismatched_freq_in.
503         (profile_record_account_profile): New function
504         that calculates time and size of a function.
505         * cfghooks.h (struct profile_record): Remove
506         all tuples.
507         (struct cfg_hooks): Remove after_pass flag.
508         (account_profile_record): Rename to ...
509         (profile_record_check_consistency): ... this.
510         (profile_record_account_profile): New.
511         * cfgrtl.c (rtl_account_profile_record): Remove
512         after_pass flag.
513         * passes.c (check_profile_consistency): Do only
514         checking.
515         (account_profile): Calculate size and time of
516         function only.
517         (pass_manager::dump_profile_report): Reformat
518         output.
519         (execute_one_ipa_transform_pass): Call
520         consistency check before clean upand call account_profile
521         after a clean up is done.
522         (execute_one_pass): Call check_profile_consistency and
523         account_profile instead of using after_pass flag..
524         * tree-cfg.c (gimple_account_profile_record): Likewise.
526 2018-11-13  Martin Liska  <mliska@suse.cz>
528         PR sanitizer/87930
529         * config/i386/i386.c (ix86_option_override_internal): Error
530         about usage -mabi=ms and -fsanitize={,kernel-}address.
532 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
534         * config/aarch64/aarch64.c
535         (aarch64_vectorize_preferred_vector_alignment): Change return type to
536         poly_uint64.
537         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
538         alignment being a poly int.
539         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
540         return type to poly_uint64.
541         * target.def (default_preferred_vector_alignment): Likewise.
542         * targhooks.c (default_preferred_vector_alignment): Likewise.
543         * targhooks.h (default_preferred_vector_alignment): Likewise.
544         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
545         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
546         poly int.
547         (vect_update_misalignment_for_peel): Likewise.
548         (vect_enhance_data_refs_alignment): Likewise.
549         (vect_find_same_alignment_drs): Likewise.
550         (vect_duplicate_ssa_name_ptr_info): Likewise.
551         (vect_setup_realignment): Likewise.
552         (vect_can_force_dr_alignment_p): Change alignment parameter type to
553         poly_uint64.
554         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
555         mask with a compile time variable vector alignment.
556         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
557         int.
558         (vect_do_peeling): Exit early if vector alignment is not constant.
559         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
560         a poly int.
561         (vectorizable_store): Likewise.
562         (vectorizable_load): Likweise.
563         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
564         poly_uint64.
565         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
566         poly int.
567         (vect_can_force_dr_alignment_p): Change alignment parameter type to
568         poly_uint64.
570 2018-11-13  Richard Biener  <rguenther@suse.de>
572         PR tree-optimization/87962
573         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
574         detect outer reduction for disqualifying in-loop uses.
576 2018-11-13  Richard Biener  <rguenther@suse.de>
578         PR tree-optimization/87967
579         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
580         for constants for the scalar loop.
582 2018-11-13  Alan Modra  <amodra@gmail.com>
584         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
585         integer constants when -mcmodel=medium.
587 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
589         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
590         * config/arc/elf.h (PROFILE_HOOK): Define.
591         * config/arc/linux.h (PROFILE_HOOK): Likewise.
593 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
595         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
596         jump table data in the text section.
598 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
600         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
601         to fit the eh_return pattern.
602         * config/arc/arc.md (eh_return): Define.
603         (VUNSPEC_ARC_EH_RETURN): Likewise.
604         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
605         new implementation.
606         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
609 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
611         * common/config/arc/arc-common.c (arc_option_optimization_table):
612         Millicode optimization is default on for size optimizations.
613         * config/arc/arc-protos.h (arc_check_multi): New function.
614         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
615         (ENTER_LEAVE_START_REG): Define.
616         (ENTER_LEAVE_END_REG): Likewise.
617         (arc_override_options): Disable millicode when long calls option
618         is on.
619         (arc_frame_info): Change it from int to bool.
620         (arc_compute_frame_size): Clean up.
621         (arc_save_restore): Remove.
622         (frame_save_reg): New function.
623         (frame_restore_reg): Likewise.
624         (arc_enter_leave_p): Likewise.
625         (arc_save_callee_saves): Likewise.
626         (arc_restore_callee_saves): Likewise.
627         (arc_save_callee_enter): Likewise.
628         (arc_restore_callee_leave): Likewise.
629         (arc_save_callee_milli): Likewise.
630         (arc_restore_callee_milli): Likewise.
631         (arc_expand_prologue): Reimplement to emit enter/leave
632         instructions.
633         (arc_expand_epilogue): Likewise.
634         (arc_check_multi): New function.
635         * config/arc/arc.md (push_multi_fp): New pattern.
636         (push_multi_fp_blink): Likewise.
637         (pop_multi_fp): Likewise.
638         (pop_multi_fp_blink): Likewise.
639         (pop_multi_fp_ret): Likewise.
640         (pop_multi_fp_blink_ret): Likewise.
641         * config/arc/arc.opt (mmillicode): Update option.
642         (mcode-density-frame): New option.
643         * config/arc/predicates.md (push_multi_operand): New predicate.
644         (pop_multi_operand): Likewise.
645         * doc/invoke.texi (ARC): Update ARC options information.
648 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
650         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
651         * config/arc/arc.c (operands_ok_ldd_std): New function.
652         (mem_ok_for_ldd_std): Likewise.
653         (gen_operands_ldd_std): Likewise.
654         * config/arc/arc.md: Add peephole2 rules for std/ldd.
656 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
658         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
659         test on stack_usage_file.
660         (lang_dependent_init): Do not open the .su file if generating LTO.
662 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
664         PR rtl-optimization/87918
665         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
666         simplify_gen_relational rather than simplify_gen_binary.
668 2018-11-13  Richard Biener  <rguenther@suse.de>
670         * tree-ssanames.h (set_range_info): Use value_range_base.
671         (get_range_info): Likewise.
672         * tree-ssanames.c (set_range_info): Likewise.
673         (get_range_info): Likewise.
674         * tree-vrp.c (value_range_base::union_helper): Split
675         out common parts of value_range[_base]::union_.
676         (value_range_base::union_): Update.
677         (value_range::union_): Likewise.
678         (determine_value_range_1): Use value_range_base.
679         (determine_value_range): Likewise.
680         * tree-vrp.h (value_range_base::union_helper): Move ...
681         (value_range::union_helper): ... from here.
683 2018-11-13  Alan Modra  <amodra@gmail.com>
685         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
686         offset for PRE_DEC.
687         (rs6000_secondary_reload_gpr): Don't call find_replacement.
689 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
691         PR middle-end/59634
692         * doc/invoke.texi (Optimize Options): Clarify that the
693         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
694         apply to data cache size.
696 2018-11-13  Alan Modra  <amodra@gmail.com>
698         * config/rs6000/predicates.md (logical_const_operand),
699         (logical_operand): Correct comment.
700         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
702 2018-11-13  Alan Modra  <amodra@gmail.com>
704         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
706 2018-11-13  Alan Modra  <amodra@gmail.com>
708         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
709         description.
710         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
711         covered by alternative.
712         (movcc_internal1): Ignore h for register preference.
713         (mov<mode>_hardfloat64): Likewise.
714         (mov<mode>_softfloat): Ignore c, l, h for register preference.
716 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
718         PR preprocessor/47823
719         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
720         documentation to...
721         (Pragmas): ...here.
722         * doc/extend.texi (Pragmas): Note additional pragmas documented
723         in the CPP manual.
725 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
727         PR middle-end/86677
728         PR middle-end/87528
729         * tree-scalar-evolution.c (expression_expensive_p): Make
730         BUILTIN POPCOUNT as expensive when backend does not define it.
732 2018-11-12  Fredrik Noring  <noring@nocrew.org>
734         * config.gcc: Update with-llsc defaults for MIPS r5900.
736 2018-11-12  Martin Liska  <mliska@suse.cz>
738         PR gcov-profile/87442
739         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
740         options.
741         * doc/invoke.texi: Document them.
742         * tree-profile.c (parse_profile_filter): New.
743         (parse_profile_file_filtering): Likewise.
744         (release_profile_file_filtering): Likewise.
745         (include_source_file_for_profile): Likewise.
746         (tree_profiling): Filter source files based on the
747         newly added options.
749 2018-11-12  Martin Liska  <mliska@suse.cz>
751         PR target/87903
752         * doc/extend.texi: Add missing values for __builtin_cpu_is and
753         __builtin_cpu_supports for x86 target.
755 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
757         PR middle-end/21110
758         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
759         have to be Pmode.
760         (Arithmetic): Likewise for "lo_sum".
762 2018-11-12  Renlin Li  <renlin.li@arm.com>
764         PR target/87815
765         * dse.c (get_stored_val): Add check for compile-time constantness
766         of gap.
768 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
770         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
771         (ARMv8_5a): New fgroup.
772         (armv8.5-a): New arch.
773         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
774         sb and predres.
775         * config/arm/arm-tables.opt: Regenerate.
776         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
777         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
778         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
779         Add matching rules for -march=armv8.5-a and extensions.
780         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
781         Add sb and predres to all armv8-a except armv8.5-a.
783 2018-11-12  Richard Biener  <rguenther@suse.de>
785         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
786         overload for single value.
787         (value_range[_base]::set_nonnull): New.
788         (value_range[_base]::set_null): Likewise.
789         (value_range): Document bitmap copying behavior, mark
790         copy constructor and assignment operator deleted.
791         (value_range::move): New.
792         (value_range::set_and_canonicalize): Default bitmap to zero.
793         (set_value_range_to_nonnull): Remove.
794         (set_value_range_to_null): Likewise.
795         (set_value_range): Likewise.
796         (set_value_range_to_value): Likewise.
797         (extract_range_from_unary_expr): Work on value_range_base.
798         (extract_range_from_binary_expr_1): Likewise.  Rename to...
799         (extract_range_from_binary_expr): ... this.
800         * tree-vrp.c (value_range::update): Clear equiv bitmap
801         if required.
802         (value_range::move): New, move equiv bitmap.
803         (value_range_base::set_undefined): Avoid assignment.
804         (value_range::set_undefined): Likewise.
805         (value_range_base::set_varying): Likewise.
806         (value_range::set_varying): Likewise.
807         (set_value_range): Remove.
808         (value_range_base::set): New overload for value.
809         (value_range::set): Likewise.
810         (set_value_range_to_nonnull): Remove.
811         (value_range_base::set_nonnull): New.
812         (value_range::set_nonnull): Likewise.
813         (set_value_range_to_null): Remove.
814         (value_range_base::set_null): New.
815         (value_range::set_null): Likewise.
816         (range_is_null): Work on value_range_base.
817         (range_is_nonnull): Likewise.
818         (ranges_from_anti_range): Likewise.
819         (extract_range_into_wide_ints): Likewise.
820         (extract_range_from_multiplicative_op): Likewise.
821         (extract_range_from_binary_expr): Likewise.  Update for API changes.
822         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
823         handling.
824         (value_range::intersect_helper): Avoid copy and assignment.
825         (value_range::union_helper): Likewise.
826         (determine_value_range_1): Adjust.
827         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
828         Avoid assignment by using move.
829         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
830         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
831         Likewise.
832         * tree-ssanames.c (get_range_info): Likewise.
833         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
834         * vr-values.c (vr_values::get_value_range): Adjust.
835         (vr_values::update_value_range): Likewise.
836         (symbolic_range_based_on_p): Work on value_range_base.
837         (vr_values::extract_range_from_binary_expr): Use value_range_base.
838         (vr_values::extract_range_from_unary_expr): Likewise.
839         (vr_values::extract_range_from_cond_expr): Avoid assignment.
840         (vr_values::extract_range_from_comparison): Adjust.
841         (vr_values::check_for_binary_op_overflow): Use value_range_base.
842         (vr_values::extract_range_basic): Adjust.
843         (vr_values::adjust_range_with_scev): Likewise.
844         (vr_values::vrp_visit_assignment_or_call): Likewise.
845         (vr_values::get_vr_for_comparison): Change API to avoid
846         assignment and copy construction.
847         (vr_values::compare_name_with_value): Adjust accordingly.
848         (vr_values::compare_names): Likewise.
849         (vr_values::extract_range_from_phi_node): Avoid assignment and
850         bogus in-place modify of equiv bitmap.
851         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
852         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
853         for extract_range_from_unary_expr API change.
854         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
856 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
858         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
860 2018-11-12  Richard Biener  <rguenther@suse.de>
862         * tree-vrp.h (value_range_base::symbolic_p,
863         value_range_base::constant_p, value_range_base::zero_p,
864         value_range_base::singleton_p): Move from value_range.
865         (value_range::dump): Add.
866         * gimple-ssa-evrp-analyze.c
867         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
868         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
869         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
870         Use set_varying.
871         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
872         (value_range::constant_p): Likewise.
873         (value_range::singleton_p): Likewise.
874         (value_range_base::dump): Add.
875         (set_value_range_to_undefined): Remove.
876         (set_value_range_to_varying): Likewise.
877         (range_int_cst_p): Take value_range_base argument.
878         (range_int_cst_singleton_p): Likewise.
879         (value_range_constant_singleton): Likewise.
880         (vrp_set_zero_nonzero_bits): Likewise.
881         (extract_range_from_multiplicative_op): Use set_varying.
882         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
883         (extract_range_from_unary_expr): Likewise.
884         (dump_value_range_base): Change to overload of dump_value_range.
885         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
886         (vrp_prop::visit_stmt): Likewise.
887         (value_range::intersect_helper): Likewise.
888         (value_range::union_helper): Likewise.
889         (determine_value_range_1): Likewise.
891 2018-11-12  Richard Biener  <rguenther@suse.de>
893         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
894         (set_value_range_to_null): Likewise.
895         * vr-values.c (vr_values::extract_range_from_comparison):
896         Clear equiv for constant singleton ranges.
898 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
900         * config/i386/sse.md: Combine VFIXUPIMM* patterns
901         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
902         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
903         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
904         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
905         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
906         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
908 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
910         PR c/69502
911         * doc/extend.texi (Common Type Attributes): For the align type
912         attribute, copy language about decreasing alignment from the
913         corresponding variable attribute.
915 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
917         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
918         -frounding-math.
920 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
922         PR c++/43105
923         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
924         -frtti and -fno-rtti code.
926 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
928         PR c/26366
929         * doc/extend.texi (Other Builtins): Document probability associated
930         with __builtin_expect.
932 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
934         PR target/87928
935         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
936         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
937         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
938         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
940 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
942         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
944 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
946         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
948 2018-11-11  Richard Biener  <rguenther@suse.de>
950         * tree-vrp.h (class value_range_base): New base class for
951         value_range containing all but the m_equiv member.
952         (dump_value_range_base): Add.
953         (range_includes_zero_p): Work on value_range_base.
954         * tree-vrp.c (value_range_base::set): Split out base handling
955         from...
956         (value_range::set): this.
957         (value_range::set_equiv): New.
958         (value_range_base::value_range_base): New constructors.
959         (value_range_base::check): Split out base handling from...
960         (value_range::check): this.
961         (value_range::equal_p): Refactor in terms of
962         ignore_equivs_equal_p which is now member of the base.
963         (value_range_base::set_undefined): New.
964         (value_range_base::set_varying): Likewise.
965         (value_range_base::dump):Split out base handling from...
966         (value_range::dump): this.
967         (value_range_base::set_and_canonicalize): Split out base handling
968         from...
969         (value_range::set_and_canonicalize): this.
970         (value_range_base::union_): New.
971         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
972         for m_vr.
973         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
974         instead of value_range everywhere.
975         (ipcp_vr_lattice::print): Use dump_value_range_base.
976         (ipcp_vr_lattice::meet_with): Adjust.
977         (ipcp_vr_lattice::meet_with_1): Likewise.
978         (ipa_vr_operation_and_type_effects): Likewise.
979         (propagate_vr_across_jump_function): Likewise.
980         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
981         (ipa_get_value_range): Likewise.
982         (ipa_set_jfunc_vr): Likewise.
983         (ipa_compute_jump_functions_for_edge): Likewise.
985 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
987         PR middle-end/65703
988         * doc/invoke.texi (Optimize Options): Add @opindex entries
989         for the positive forms of -fno-xxx and -mno-xxx options
990         that were lacking them.
992 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
994         * combine.c (make_more_copies): Only make an intermediate copy if the
995         dest of a move is a pseudo.
997 2018-11-09  Maya Rashish  <coypu@sdf.org>
999         PR target/87221
1000         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1001         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1003 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1005         PR driver/41179
1006         PR middle-end/65703
1007         * doc/invoke.texi (Optimize Options): Clarify default behavior
1008         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1010 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1012         PR tree-optimization/87940
1013         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1014         Fold PLUS_EXPR correctly.
1016 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1018         PR target/87762
1019         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1020         (annotate_constant_pool_refs): Skip insns which support
1021         relative addressing.
1022         (annotate_constant_pool_refs_1): New helper function.
1023         (find_constant_pool_ref): Skip insns which support relative
1024         addression.
1025         (find_constant_pool_ref_1): New helper function.
1026         (replace_constant_pool_ref): Skip insns which support
1027         relative addressing.
1028         (replace_constant_pool_ref_1): New helper function.
1029         (s390_mainpool_start): Adapt to the new signature.
1030         (s390_mainpool_finish): Likewise.
1031         (s390_chunkify_start): Likewise.
1032         (s390_chunkify_finish): Likewise.
1033         (pass_s390_early_mach::execute): Likewise.
1034         (s390_prologue_plus_offset): Likewise.
1035         (s390_emit_prologue): Likewise.
1036         (s390_emit_epilogue): Likewise.
1038 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1040         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1041         but unsupported lastprivate with conditional modifier.
1043 2018-11-09  Jeff Law  <law@redhat.com>
1045         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1046         unused argument better.  Add gcc_unreachable to silence warning.
1048 2018-11-09  Martin Sebor  <msebor@redhat.com>
1050         PR middle-end/81824
1051         * attribs.c (has_attribute): New helper function.
1052         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1053         * attribs.h (decls_mismatched_attributes): Declare.
1054         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1055         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1056         * common.opt (-Wattribute-alias): Take an argument.
1057         (-Wno-attribute-alias): New option.
1058         * doc/extend.texi (Common Function Attributes): Document copy.
1059         (Common Variable Attributes): Same.
1060         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1061         (-Wattribute-alias): Document new option argument.
1063 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1065         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1066         matching criteria.  Remove unused array initializer.
1068 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1069             Jinsong Ji  <jji@us.ibm.com>
1071         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1072         constraints by introducing a new temporary.
1073         (_mm_cvtss_si64): Likewise.
1075 2018-11-09  Martin Liska  <mliska@suse.cz>
1077         * common.opt: Add -fipa-stack-alignment flag.
1078         * doc/invoke.texi: Document it.
1079         * final.c (rest_of_clean_state): Guard stack
1080         shrinking with flag.
1082 2018-11-09  Martin Liska  <mliska@suse.cz>
1084         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1085         to ...
1086         (ipa_discover_variable_flags): ... this.
1087         * common.opt: Come up with new flag -fipa-reference-addressable.
1088         * doc/invoke.texi: Document it.
1089         * ipa-reference.c (propagate): Call the renamed fn.
1090         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1091         Likewise.
1092         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1093         ...
1094         (ipa_discover_variable_flags): ... this.  Discover
1095         non-addressable variables only with the newly added flag.
1096         * opts.c: Enable the newly added flag with -O1 and higher
1097         optimization level.
1099 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1101         * json.cc (selftest::test_writing_literals): Fix comment.
1103 2018-11-09  Martin Liska  <mliska@suse.cz>
1105         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1106         string to a stack buffer.
1107         (aarch64_parse_cpu): Likewise.
1108         (aarch64_parse_tune): Likewise.
1110 2018-11-09  Richard Biener  <rguenther@suse.de>
1112         PR tree-optimization/87953
1113         * tree-vect-loop.c (vectorizable_reduction): For analysis
1114         always pass ops[0] to vectorizable_condition.
1116 2018-11-09  Stafford Horne  <shorne@gmail.com>
1117             Richard Henderson  <rth@twiddle.net>
1118             Joel Sherrill  <joel@rtems.org>
1120         * common/config/or1k/or1k-common.c: New file.
1121         * config/or1k/*: New.
1122         * config.gcc (or1k*-*-*): New.
1123         * configure.ac (or1k*-*-*): New test for openrisc tls.
1124         * configure: Regenerated.
1125         * doc/install.texi: Document OpenRISC triplets.
1126         * doc/invoke.texi: Document OpenRISC arguments.
1127         * doc/md.texi: Document OpenRISC.
1129 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1131         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1132         (arm7tdmi-s): Delete CPU.
1133         (arm710t): Add aliases for arm720t and arm740t.
1134         (arm720t, arm740t): Delete CPUs.
1135         (arm920t): Add aliases for arm920, arm922t and arm940t.
1136         (arm920, arm922t, arm940t): Delete CPUs.
1137         (arm10tdmi): Add alias for arm1020t.
1138         (arm1020t): Delete CPU.
1139         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1140         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1141         (arm10e): Add aliases for arm1020e and arm1022e.
1142         (arm1020e, arm1022e): Delete CPU.
1143         * config/arm/arm.md (generic_sched): Remove entries that are now
1144         handled by aliases.
1145         (generic_vfp): Likewise.
1146         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1147         changes.
1148         * config/arm/arm-tune.md: Regenerated.
1149         * config/arm/arm-tables.opt: Regenerated.
1151 2018-11-09  Richard Biener  <rguenther@suse.de>
1153         PR tree-optimization/87621
1154         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1155         op with only phi inputs.
1156         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1157         (ch_base::copy_headers): Run CSE on copied loop headers.
1158         (pass_ch_vect::process_loop_p): Simplify.
1160 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1162         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1163         for non-w64 x86_64 biarch.
1165 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1167         PR rtl-optimization/86438
1168         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1169         of in_b for the compare if in_b is SET_DEST.
1171         PR target/87793
1172         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1173         non-toplevel UNSPEC.
1175 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1177         * tree-vrp.c (value_range::check): Do not access internals
1178         directly.
1179         (value_range::singleton_p): Same.
1180         (value_range::type): Same.
1181         (vrp_finalize): Use value_range API.
1183 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1185         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1187 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1189         * vr-values.c (vr_values::get_value_range): Use value_range API
1190         instead of piecing together ranges.
1191         (vr_values::update_value_range): Same.
1193 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1195         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1196         of performing ad-hoc calculations.
1197         * tree-ssanames.c (set_range_info): New overloaded function
1198         accepting value_range &.
1199         (get_range_info): Same.
1200         * tree-ssanames.h (set_range_info_raw): Remove.
1201         (set_range_info): New prototype.
1202         (get_range_info): Same.
1203         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1204         * tree-vrp.c (value_range::null_p): Same.
1206 2018-11-09  Jan Hubicka  <jh@suse.cz>
1208         * tree.c (fld_type_variant_equal_p): Test user align flag.
1209         (flt_type_variant): Copy user align flag.
1210         (fld_incomplete_type_of): Clear it.
1212 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1214         * config/arm/neon.md (div<mode>3): New pattern.
1216 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1218         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1219         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1220         (ix86_handle_option): Handle OPT_mptwrite.
1221         * config/i386/cpuid.h (bit_PTWRITE): Add.
1222         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1223         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1224         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1225         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1226         (ix86_option_override_internal): Handle PTA_PTWRITE.
1227         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1228         (def_builtin2): Force UINT64 to be 64bit only.
1229         * config/i386/i386.h (TARGET_PTWRITE): Add.
1230         (TARGET_PTWRITE_P): Add.
1231         (PTA_PTWRITE): Add.
1232         * config/i386/i386.md: Define ptwrite.
1233         * config/i386/i386.opt: Add -mptwrite.
1234         * config/i386/immintrin.h (_ptwrite64): Add.
1235         (_ptwrite32): Add
1236         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1237         * doc/invoke.texi: Document -mptwrite.
1239 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1241         PR rtl-optimization/87600
1242         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1243         * lra-constraints.c (process_alt_operands): Skip illegal hard
1244         register usage.  Prefer reloading non hard register operands.
1246 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1248         PR other/36572
1249         * doc/invoke.texi (Optimize Options): Clarify default behavior
1250         for -fno-sched-interblock and -fno-sched-spec.
1252 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1254         * collect2.c (linker_select):  Add USE_LLD_LD.
1255         (ld_suffixes): Add ld.lld.
1256         (main): Handle -fuse-ld=lld.
1257         * common.opt (-fuse-ld=lld): New option.
1258         * doc/invoke.texi (-fuse-ld=lld): Document.
1259         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1261 2018-11-08  Paul Koning  <ni1d@arrl.net>
1263         * config/pdp11/constraints.md: Add "Z" series constraints for use
1264         with pre-dec and post-inc addressing.
1265         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1266         (pdp11_expand_operands): Add int argument (word count).
1267         (pdp11_sp_frame_offset): Delete.
1268         (pdp11_cmp_length): New function.
1269         (pushpop_regeq): New function.
1270         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1271         Add hook.
1272         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1273         frame layout.
1274         (pdp11_initial_elimination_offset): Ditto.
1275         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1276         (output_move_multiple): Change how pointer adjustment is done.
1277         (pdp11_gen_int_label): Correct format.
1278         (output_ascii): Ditto.
1279         (pdp11_asm_output_var): Add code for DEC assembler case.
1280         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1281         value.
1282         (legitimate_const_double_p): Ditto.
1283         (pdp11_register_move_cost): Adjust for new register classes.
1284         (pdp11_regno_reg_class): Ditto.
1285         (expand_block_move): Delete.
1286         (pushpop_regeq): New function.
1287         (pdp11_legitimate_address_p): Bugfix in check for constant
1288         offset.
1289         (pdp11_sp_frame_offset): Delete.
1290         (pdp11_reg_save_size): New helper function for new frame layout.
1291         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1292         (pdp11_expand_shift): Bugfix in check for constant shift count.
1293         (pdp11_shift_length): Ditto.
1294         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1295         (pdp11_cmp_length): New function.
1296         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1297         some compile options.
1298         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1299         (CALL_USED_REGISTERS): Ditto.
1300         (ELIMINABLE_REGS): Ditto.
1301         (REGISTER_NAMES): Ditto.
1302         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1303         constraints.
1304         (REG_CLASS_NAMES): Ditto.
1305         (REG_CLASS_CONTENTS): Ditto.  Also remove
1306         HARD_FRAME_POINTER_REGNUM.
1307         (CPU_REG_CLASS): New macro.
1308         (CLASS_MAX_NREGS): Adjust for new register classes.
1309         (FUNCTION_PROFILER): Make no-op.
1310         (may_call_alloca): Remove unused declaration.
1311         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1312         (ASM_OUTPUT_SKIP): Fix format.
1313         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1314         (HARD_FRAME_POINTER_REGNUM): Remove.
1315         (return): Delete.
1316         (*rts): Rename.  Remove epilogue related checks.
1317         (cmpsi, cmpdi): New insn.
1318         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1319         (mov<mode>): Change constraints to enforce that push/pop
1320         destination cannot use the same register as source.
1321         (*mov<mode><cc_cc>): Ditto.
1322         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1323         at assembly output rather than as RTL expander.
1324         (zero_extendqihi2): Bugfix in check for same registers.
1325         (adddi3_nocc): Bugfix in check for constant operand.
1326         (addsi3_nocc): Ditto.
1327         (subdi3_nocc): Ditto.
1328         (subsi3_nocc): Ditto.
1329         (negdi2_nocc): Copy input to pdp11_expand_operands.
1330         (negsi2_nocc): Ditto.
1331         (bswap2_nocc): Ditto.
1332         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1333         * config/pdp11/t-pdp11: Use -Os.
1335 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1337         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1338         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1339         (gen_comm_data): Print out CPU alias tables.
1340         (check_cpu): Match aliases when checking the CPU name.
1341         * config/arm/arm-protos.h (cpu_alias): New structure.
1342         (cpu_option): Add entry for aliases.
1343         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1344         strongarm1100 and strongarm1110.
1345         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1346         (config/arm/arm-generic.md): Remove redundant references to
1347         strongarm110, strongarm1100 and strongarm1110.
1348         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1349         Scan aliases for additional hints.
1350         (arm_parse_cpu_option_name): Also match a cpu name against the list
1351         of aliases.
1352         * config/arm/arm-tables.opt: Regenerated.
1353         * config/arm/arm-tune.md: Regenerated.
1355 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1357         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1358         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1359         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1360         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1361         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1362         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1363         * gengtype.c (open_base_files): Add omp-general.h.
1364         * gimple.c (gimple_build_omp_critical):
1365         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1366         gimple_omp_taskgroup_set_clauses.
1367         (gimple_build_omp_atomic_load): Add mo argument, call
1368         gimple_omp_atomic_set_memory_order.
1369         (gimple_build_omp_atomic_store): Likewise.
1370         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1371         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1372         instead of GSS_OMP.
1373         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1374         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1375         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1376         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1377         different value for GF_OMP_ATOMIC_NEED_VALUE.
1378         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1379         comments.
1380         (struct gimple_statement_omp_single_layout): And remove here.
1381         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1382         than gimple_statement_omp_single_layout.
1383         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1384         GIMPLE_OMP_TEAMS.
1385         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1386         (gimple_omp_subcode): Formatting fix.
1387         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1388         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1389         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1390         gimple_omp_teams_host, gimple_omp_teams_set_host,
1391         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1392         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1393         gimple_omp_taskgroup_set_clauses): New inline functions.
1394         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1395         (gimple_build_omp_atomic_store): Likewise.
1396         (gimple_omp_atomic_seq_cst_p): Remove.
1397         (gimple_omp_atomic_memory_order): New function.
1398         (gimple_omp_atomic_set_seq_cst): Remove.
1399         (gimple_omp_atomic_set_memory_order): New function.
1400         (gimple_build_omp_taskgroup): Add clauses argument.
1401         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1402         (dump_gimple_omp_task): Print taskwait with depend clauses.
1403         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1404         dump_omp_atomic_memory_order.
1405         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1406         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1407         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1408         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1409         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1410         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1411         ORT_UNTIED_TASKLOOP enumerators.
1412         (enum gimplify_defaultmap_kind): New.
1413         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1414         target_map_pointers_as_0len_arrays members, add defaultmap.
1415         (new_omp_context): Initialize defaultmap member.
1416         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1417         (maybe_fold_stmt): Don't fold even in host teams regions.
1418         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1419         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1420         ctx->omp_firstprivatize_variable.
1421         (omp_add_variable): Don't add private/firstprivate for VLAs in
1422         ORT_TASKGROUP.
1423         (omp_default_clause): Print "taskloop" rather than "task" if
1424         ORT_*TASKLOOP.
1425         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1426         Handle new defaultmap clause kinds.
1427         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1428         iterator to be lastprivate or private.  Fix up diagnostics if linear
1429         is used on collapse>1 simd iterator.
1430         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1431         (gimplify_omp_depend): New function.
1432         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1433         combined parallel master taskloop{, simd} if taskloop has
1434         firstprivate, lastprivate or reduction clause.  Handle
1435         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1436         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1437         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1438         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1439         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1440         lastprivate.
1441         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1442         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1443         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1444         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1445         (gimplify_omp_task): Handle taskwait with depend clauses.
1446         (gimplify_omp_for): Add shared clause on parallel for combined
1447         parallel master taskloop{, simd} if taskloop has firstprivate,
1448         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1449         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1450         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1451         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1452         __for_end and __for_range temporaries on OMP_PARALLEL for
1453         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1454         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1455         sandwiched in between two taskloops.
1456         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1457         instead of ctx->omp_firstprivatize_variable.
1458         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1459         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1460         host teams, use gimplify_and_return_first etc. for body like
1461         for target or target data constructs, and at the end call
1462         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1463         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1464         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1465         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1466         gimple_omp_atomic_set_seq_cst calls.
1467         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1468         case, handle taskgroup clauses.
1469         * lto-streamer-out.c (hash_tree): Handle
1470         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1471         * Makefile.in (GTFILES): Add omp-general.h.
1472         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1473         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1474         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1475         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1476         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1477         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1478         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1479         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1480         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1481         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1482         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1483         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1484         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1485         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1486         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1487         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1488         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1489         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1490         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1491         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1492         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1493         non-worksharing loops.
1494         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1495         (determine_parallel_type): Don't combine parallel with worksharing
1496         which has _reductemp_ clause.
1497         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1498         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1499         if there is nonmonotonic modifier or if there is no modifier and no
1500         ordered clause.  For dynamic and guided schedule without monotonic
1501         and nonmonotonic modifier, default to nonmonotonic.
1502         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1503         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1504         task reductions.
1505         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1506         there are any reduction clauses.
1507         (expand_taskwait_call): New function.
1508         (expand_teams_call): New function.
1509         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1510         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1511         depend clauses.
1512         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1513         of worksharing loops with task reductions.
1514         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1515         expansion of worksharing loops with task reductions.
1516         (expand_omp_sections): Handle expansion of sections with task
1517         reductions.
1518         (expand_omp_synch): For host teams call expand_omp_taskreg.
1519         (omp_memory_order_to_memmodel): New function.
1520         (expand_omp_atomic_load, expand_omp_atomic_store,
1521         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1522         instead of gimple_omp_atomic_seq_cst_p.
1523         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1524         depend clauses as a standalone directive.
1525         * omp-general.c (enum omp_requires): New variable.
1526         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1527         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1528         GT_EXPR loops depending on incr sign.  Formatting fixes.
1529         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1530         (enum omp_requires): New enum.
1531         (omp_requires_mask): Declare.
1532         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1533         Fix comment typos.
1534         * omp-low.c (struct omp_context): Add task_reductions and
1535         task_reduction_map fields.
1536         (is_host_teams_ctx): New function.
1537         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1538         (use_pointer_for_field): Use is_global_var instead of
1539         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1540         in outer contexts.
1541         (build_outer_var_ref): Ignore taskgroup outer contexts.
1542         (delete_omp_context): Release task_reductions and task_reduction_map.
1543         (scan_sharing_clauses): Don't add any fields for reduction clause on
1544         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1545         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1546         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1547         Handle OMP_CLAUSE_NONTEMPORAL.
1548         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1549         needed.
1550         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1551         clauses with task modifier.
1552         (scan_omp_task): Handle taskwait with depend clauses.
1553         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1554         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1555         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1556         (scan_omp_for): Fix comment formatting.
1557         (scan_omp_teams): Handle host teams constructs.
1558         (check_omp_nesting_restrictions): Allow teams with no outer
1559         OpenMP context.  Adjust diagnostics for teams strictly nested into
1560         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1561         inside of simd regions.
1562         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1563         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1564         taskreg_nesting_level while scanning host teams construct.
1565         (task_reduction_read): New function.
1566         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1567         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1568         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1569         second argument create_tmp_var if it is NULL.  Don't ignore shared
1570         clauses in is_host_teams_ctx contexts.  Handle
1571         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1572         clauses.
1573         (lower_reduction_clauses): Ignore reduction clauses with task
1574         modifier.  Remove second argument create_tmp_var if it is NULL.
1575         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1576         (lower_send_clauses): Ignore reduction clauses with task modifier.
1577         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1578         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1579         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1580         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1581         contexts.
1582         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1583         lower_omp_task_reductions): New functions.
1584         (lower_omp_sections): Handle reduction clauses with taskgroup
1585         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1586         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1587         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1588         modifiers.
1589         (lower_omp_taskgroup): Handle taskgroup reductions.
1590         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1591         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1592         (lower_depend_clauses): If there are any
1593         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1594         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1595         seen, assume lowering is done already and return early.  Set kind
1596         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1597         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1598         parallel construct.  Handle reduction clause on taskloop construct.
1599         Handle taskwait with depend clauses.
1600         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1601         for host teams constructs.
1602         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1603         nontemporal and _reductemp_ clause entries.
1604         (omp_clause_code_name): Likewise.
1605         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1606         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1607         * tree-core.h (enum omp_clause_code): Add
1608         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1609         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1610         (struct tree_base): Add omp_atomic_memory_order field into union.
1611         Remove OMP_ATOMIC_SEQ_CST comment.
1612         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1613         and OMP_CLAUSE_DEPEND_DEPOBJ.
1614         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1615         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1616         OpenMP constructs with body and clauses operands.
1617         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1618         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1619         (OMP_TASKGROUP_CLAUSES): Define.
1620         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1621         OMP_CLAUSE__LOOPTEMP_.
1622         (OMP_ATOMIC_SEQ_CST): Remove.
1623         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1624         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1625         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1626         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1627         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1628         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1629         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1630         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1631         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1632         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1633         Define.
1634         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1635         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1636         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1637         (convert_local_omp_clauses): Likewise.  Remove useless test.
1638         * tree-parloops.c (create_call_for_reduction_1): Pass
1639         OMP_MEMORY_ORDER_RELAXED as new argument to
1640         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1641         * tree-pretty-print.c (dump_omp_iterators): New function.
1642         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1643         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1644         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1645         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1646         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1647         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1648         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1649         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1650         (dump_omp_atomic_memory_order): New function.
1651         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1652         taskwait with depend clauses.
1653         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1654         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1655         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1656         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1657         write_ts_omp_clause_tree_pointers): Likewise.
1659 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1661         PR ipa/86395
1662         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1663         "-missed", "-note", and "-all" sub-options.
1664         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1665         API.
1666         (can_early_inline_edge_p): Likewise.
1667         (want_early_inline_function_p): Likewise.
1668         (want_inline_self_recursive_call_p): Likewise.
1669         (recursive_inlining): Likewise.
1670         (inline_small_functions): Likewise.
1671         (flatten_function): Likewise.
1672         (ipa_inline): Likewise.
1673         (inline_always_inline_functions): Likewise.
1674         (early_inline_small_functions): Likewise.
1675         (early_inliner): Likewise.
1676         * tree-inline.c (expand_call_inline): Likewise.
1678 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1680         * pretty-print.c (pp_format): Handle %f.
1681         (selftest::test_pp_format): Add test of %f.
1682         * pretty-print.h (pp_double): New macro.
1684 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1686         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1687         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1688         (dump_pretty_printer::decode_format): Implement "%C" for
1689         cgraph_node *.
1690         (selftest::test_capture_of_dump_calls): Rename "where" to
1691         "stmt_loc".  Convert test_decl to a function decl and set its
1692         location.  Add a symbol_table_test RAII instance and a
1693         cgraph_node, using it to test "%C" and dump_symtab_node.
1695 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1697         PR middle-end/87916
1698         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1700 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1702         * cgraph.c: Include "selftest.h".
1703         (saved_symtab): New variable.
1704         (selftest::symbol_table_test::symbol_table_test): New ctor.
1705         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1706         (selftest::test_symbol_table_test): New test.
1707         (selftest::cgraph_c_tests): New.
1708         * cgraph.h (saved_symtab): New decl.
1709         (selftest::symbol_table_test): New class.
1710         * selftest-run-tests.c (selftest::run_tests): Call
1711         selftest::cgraph_c_tests.
1712         * selftest.h (selftest::cgraph_c_tests): New decl.
1714 2018-11-08  Richard Biener  <rguenther@suse.de>
1716         * tree-data-ref.h (lambda_int): New typedef.
1717         (lambda_vector_gcd): Adjust.
1718         (lambda_vector_new): Likewise.
1719         (lambda_matrix_new): Likewise.
1720         * tree-data-ref.c  (print_lambda_vector): Adjust.
1722 2018-11-08  Richard Biener  <rguenther@suse.de>
1724         PR tree-optimization/87929
1725         * tree-complex.c (expand_complex_comparison): Clean EH.
1727 2018-11-08  Martin Liska  <mliska@suse.cz>
1729         * doc/extend.texi: Reword.
1730         * predict.c (expr_expected_value_1): Likewise.
1732 2018-11-08  Richard Biener  <rguenther@suse.de>
1734         PR tree-optimization/87913
1735         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1736         of extreme values to ordered comparisons.
1738 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1740         PR middle-end/42726
1741         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1743 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1745         * doc/invoke.texi: Remove leading dash from @opindex entries
1746         throughout the file.
1748 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1750         PR driver/80828
1751         * doc/invoke.texi (Option Summary): Add -e and --entry.
1752         (Link Options): Likewise.
1754 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1756         PR 87926
1757         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1758         --disable-checking bootstrap.
1760 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1762         * configure: Regenerated.
1764 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1766         PR c/87691
1767         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1768         to the mode of the widest field iff the widest field has mode class
1769         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1770         reference.
1772 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1774         * common.opt: Add -Wattribute-warning.
1775         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1776         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1777         call to allow user configure behavior of "warning" attribute.
1779 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1781         * target.def: Put @: after every vs., e.g., and i.e. where it is
1782         followed by whitespace.
1783         * doc/extend.texi: Ditto.
1784         * doc/fragments.texi: Ditto.
1785         * doc/gimple.texi: Ditto.
1786         * doc/implement-c.texi: Ditto.
1787         * doc/install.texi: Ditto.
1788         * doc/invoke.texi: Ditto.
1789         * doc/md.texi: Ditto.
1790         * doc/plugins.texi: Ditto.
1791         * doc/rtl.texi: Ditto.
1792         * doc/sourcebuild.texi: Ditto.
1793         * doc/tm.texi.in: Ditto.
1794         * doc/ux.texi: Ditto.
1795         * doc/tm.texi: Regenerate.
1797 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1799         * config/arm/arm-cpus.in (ares): New entry.
1800         * config/arm/arm-tables.opt: Regenerate.
1801         * config/arm/arm-tune.md: Likewise.
1802         * doc/invoke.texi (ARM Options): Document ares.
1804 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1806         * config/aarch64/aarch64-cores.def (ares): Define.
1807         * config/aarch64/aarch64-tune.md: Regenerate.
1808         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1810 2018-11-07  Jan Hubicka  <jh@suse.cz>
1812         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1813         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1814         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1815         TYPE_DECL.
1817 2018-11-07  Richard Biener  <rguenther@suse.de>
1819         PR tree-optimization/87914
1820         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1821         of nested cycles.
1822         (vectorizable_reduction): Handle shifts and rotates by dispatching
1823         to vectorizable_shift.
1824         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1825         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1826         def cases.
1827         (vectorizable_shift): Export and handle being called as
1828         vect_nested_cycle.
1829         (vect_analyze_stmt): Call vectorizable_shift after
1830         vectorizable_reduction.
1831         * tree-vectorizer.h (vectorizable_shift): Declare.
1833 2018-11-07  Jan Hubicka  <jh@suse.cz>
1835         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1836         than const decls in TREE_VALUE of enum.
1837         (dump_type_inheritance_graph): Improve duplicate dumping.
1838         (free_enum_values): New.
1839         (build_type_inheritance_graph): Use it.
1840         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1841         which are not main variants or not ODR types.
1842         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1844 2018-11-07  Richard Biener  <rguenther@suse.de>
1846         * ipa-inline.c (want_inline_small_function_p): Compute
1847         big_speedup_p lazily and last.
1849 2018-11-07  Jan Hubicka  <jh@suse.cz>
1851         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1852         building incomplete variant of complete type.
1853         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1854         variant of complete type.
1856 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1858         * config/mips/mips.c: Fix typo in documentation of
1859         mips_loongson_ext2_prefetch_cookie.
1860         (mips_option_override): fix brain twister logical.
1861         * config/mips/mips.h: Fix typo in documentation of
1862         ISA_HAS_CTZ_CTO and define pattern.
1863         * config/mips/mips.md (prefetch): Hoist EXT2 above
1864         the 2EF/EXT block.
1865         (prefetch_indexed): Hoist EXT2 above the EXT block.
1867 2018-11-07  Jan Hubicka  <jh@suse.cz>
1869         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1870         return and parameter types of function and method types.
1871         (free_lang_data_in_cgraph): Update.
1873 2018-11-07  Martin Liska  <mliska@suse.cz>
1875         PR rtl-optimization/87868
1876         * postreload-gcse.c (eliminate_partially_redundant_load): Set
1877         threshold to max_count if we would overflow.
1878         * profile-count.h: Make max_count a public constant.
1880 2018-11-07  Martin Liska  <mliska@suse.cz>
1882         * mem-stats.h: Fix GNU coding style.
1884 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1886         * config/mips/gs264e.md: New.
1887         * config/mips/mips-cpus.def: Define gs264e.
1888         * config/mips/mips-tables.opt: Regenerate.
1889         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1890         gs264e.
1891         (mips_issue_rate): Add support for gs264e.
1892         (mips_multipass_dfa_lookahead): Likewise.
1893         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1894         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1895         (MIPS_ASE_MSA_SPEC): New.
1896         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1897         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1898         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1899         * config/mips/mips.md: Include gs264e.md.
1900         (processor): Add gs264e.
1901         * config/mips/mips.opt (MSA): Use Mask instead of Var.
1902         * doc/invoke.texi: Add gs264e to supported architectures.
1904 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1906         * config/mips/gs464e.md: New.
1907         * config/mips/mips-cpus.def: Define gs464e.
1908         * config/mips/mips-tables.opt: Regenerate.
1909         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1910         gs464e.
1911         (mips_issue_rate): Add support for gs464e.
1912         (mips_multipass_dfa_lookahead): Likewise.
1913         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1914         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1915         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1916         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1917         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1918         * config/mips/mips.md: Include gs464e.md.
1919         (processor): Add gs464e.
1920         * doc/invoke.texi: Add gs464e to supported architectures.
1922 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1924         * config/mips/loongson3a.md: Rename to ...
1925         * config/mips/gs464.md: ... here.
1926         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1927         as an alias of gs464 processor.
1928         * config/mips/mips-tables.opt: Regenerate.
1929         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1930         instead of PROCESSOR_LOONGSON_3A.
1931         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1932         TUNE_LOONGSON_3A.
1933         (mips_option_override): Enable MMI and EXT for gs464.
1934         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1935         Rename TUNE_LOONGSON_3A to TUNE_GS464.
1936         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1937         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1938         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1939         TARGET_LOONGSON_3A.
1940         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1941         (processor): Add gs464;
1942         * doc/invoke.texi: Add gs464 to supported architectures.
1944 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1946         * config/mips/mips-protos.h
1947         (mips_loongson_ext2_prefetch_cookie): New prototype.
1948         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1949         (mips_option_override): Enable TARGET_LOONGSON_EXT when
1950         TARGET_LOONGSON_EXT2 is true.
1951         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1952         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1953         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1954         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1955         TARGET_LOONGSON_EXT2.
1956         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1957         (define_insn "ctz<mode>2"): New insn pattern.
1958         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1959         (define_insn "prefetch_indexed_<mode>"): Include
1960         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1961         * config/mips/mips.opt (-mloongson-ext2): Add option.
1962         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1964 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1966         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1967         __mips_loongson_ext.
1968         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1969         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1970         -mloongson-ext.
1971         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1972         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1973         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1974         instead of TARGET_LOONGSON_3A.
1975         * config/mips/mips.opt (-mloongson-ext): Add option.
1976         * gcc/doc/invoke.texi (-mloongson-ext): Document.
1978 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1980         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1981         * config/mips/loongson.md: Move to ...
1982         * config/mips/loongson-mmi.md: here; Adjustment.
1983         * config/mips/loongson.h: Move to ...
1984         State as deprecated. Include loongson-mmiintrin.h for back
1985         compatibility and warning.
1986         * config/mips/loongson-mmiintrin.h: ... here.
1987         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1988         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1989         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1990         (mips_option_override): Make sure MMI use hard float;
1991         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1992         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1993         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1994         TARGET_LOONGSON_VECTORS.
1995         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1996         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1997         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1998         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1999         -mloongson-mmi.
2000         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2001         TARGET_LOONGSON_VECTORS.
2002         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2003         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2004         (Loongson MMI patterns): Include loongson-mmi.md instead of
2005         loongson.md.
2006         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2007         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2009 2018-11-07  Richard Biener  <rguenther@suse.de>
2011         PR lto/87906
2012         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2013         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2015 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2017         PR rtl-optimization/87874
2018         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2019         const wide ints.
2021 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2023         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2024         if not in indexed or indirect form.
2025         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2026         (bswap<mode>2_store): Ditto.
2028 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2030         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2031         the UNSPEC.
2033 2018-11-06  Richard Biener  <rguenther@suse.de>
2035         PR tree-optimization/86850
2036         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2037         instead of src.m_vec.
2039 2018-11-06  Jan Hubicka  <jh@suse.cz>
2041         * tree.c (fld_simplified_type_name): Break out form ...
2042         (free_lang_data_in_type): ... here.
2043         (fld_type_variant_equal_p): Use it.
2045 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2047         * config/default-d.c: Include memmodel.h.
2049         * config/sol2-d.c: New file.
2050         * config/t-sol2 (sol2-d.o): New rule.
2051         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2052         target_has_targetdm.
2054 2018-11-06  Jan Hubicka  <jh@suse.cz>
2056         * tree.c (fld_type_variant): Also copy alignment; be sure that
2057         new variant is equal.
2059 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2061         PR target/87762
2062         * config/s390/s390.md: Add relative_long attribute.
2064 2018-11-06  Jan Hubicka  <jh@suse.cz>
2066         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2067         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2068         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2069         stream TYPE_NEEDS_CONSTRUCTING.
2070         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2071         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2073 2018-11-06  Richard Biener  <rguenther@suse.de>
2075         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2076         dump-scope ...
2077         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2079 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2081         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2082         REG_CLASS_CONTENTS[GEN_REGS].
2083         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2085 2018-11-06  Jan Hubicka  <jh@suse.cz>
2087          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2089 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2091         PR sanitizer/80953
2092         * config/sol2.h (ASAN_CC1_SPEC): Define.
2093         (LD_WHOLE_ARCHIVE_OPTION): Define.
2094         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2095         (ASAN_REJECT_SPEC): Provide default.
2096         (LIBASAN_EARLY_SPEC): Define.
2097         (LIBTSAN_EARLY_SPEC): Define.
2098         (LIBLSAN_EARLY_SPEC): Define.
2099         * config/i386/sol2.h (CC1_SPEC): Redefine.
2100         (ASAN_REJECT_SPEC): Define.
2102         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2103         (TARGET_ASAN_SHADOW_OFFSET): Define.
2104         (sparc_asan_shadow_offset): New function.
2105         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2106         (ASAN_REJECT_SPEC): Define.
2108 2018-11-06  Jan Hubicka  <jh@suse.cz>
2110         * tree.c (fld_type_variant): Copy canonical type.
2111         (fld_incomplete_type_of): Check that canonical types looks sane;
2112         copy canonical type.
2113         (verify_type): Accept when incomplete type has complete canonical type.
2115 2018-11-06  Jan Hubicka  <jh@suse.cz>
2117         * tree.c (free_lang_data): Reset overwite_assembler_name,
2118         print_xnode, print_decl, print_type and print_identifier of
2119         langhooks.
2121 2018-11-06  Richard Biener  <rguenther@suse.de>
2123         PR tree-optimization/87889
2124         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2125         Do nothing if old and new arg are the same
2127 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2129         PR target/87723
2130         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2131         attributes for operands 3 and 4.
2133 2018-11-06  Richard Biener  <rguenther@suse.de>
2135         PR middle-end/18041
2136         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2137         matching bitfield insertion.
2139 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2141         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2142         (reg_next_debug_use): New.
2143         (attempt_change): Propagate adjusted expression into affected
2144         debug insns.
2145         (merge_in_block): Track uses in debug insns.
2146         (pass_inc_dec::execute): Allocate and release
2147         reg_next_debug_use.
2149 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2151         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2152         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2153         (_mm512_maskz_fixupimm_round_pd): Ditto.
2154         (_mm512_fixupimm_round_ps): Ditto.
2155         (_mm512_maskz_fixupimm_round_ps): Ditto.
2156         (_mm_fixupimm_round_sd): Ditto.
2157         (_mm_maskz_fixupimm_round_sd): Ditto.
2158         (_mm_fixupimm_round_ss): Ditto.
2159         (_mm_maskz_fixupimm_round_ss): Ditto.
2160         (_mm512_fixupimm_pd): Ditto.
2161         (_mm512_maskz_fixupimm_pd): Ditto.
2162         (_mm512_fixupimm_ps): Ditto.
2163         (_mm512_maskz_fixupimm_ps): Ditto.
2164         (_mm_fixupimm_sd): Ditto.
2165         (_mm_maskz_fixupimm_sd): Ditto.
2166         (_mm_fixupimm_ss): Ditto.
2167         (_mm_maskz_fixupimm_ss): Ditto.
2168         (_mm512_mask_fixupimm_round_pd): Update builtin.
2169         (_mm512_mask_fixupimm_round_ps): Ditto.
2170         (_mm_mask_fixupimm_round_sd): Ditto.
2171         (_mm_mask_fixupimm_round_ss): Ditto.
2172         (_mm512_mask_fixupimm_pd): Ditto.
2173         (_mm512_mask_fixupimm_ps): Ditto.
2174         (_mm_mask_fixupimm_sd): Ditto.
2175         (_mm_mask_fixupimm_ss): Ditto.
2176         * config/i386/avx512vlintrin.h:
2177         (_mm256_fixupimm_pd): Update parameters and builtin.
2178         (_mm256_maskz_fixupimm_pd): Ditto.
2179         (_mm256_fixupimm_ps): Ditto.
2180         (_mm256_maskz_fixupimm_ps): Ditto.
2181         (_mm_fixupimm_pd): Ditto.
2182         (_mm_maskz_fixupimm_pd): Ditto.
2183         (_mm_fixupimm_ps): Ditto.
2184         (_mm_maskz_fixupimm_ps): Ditto.
2185         (_mm256_mask_fixupimm_pd): Update builtin.
2186         (_mm256_mask_fixupimm_ps): Ditto.
2187         (_mm_mask_fixupimm_pd): Ditto.
2188         (_mm_mask_fixupimm_ps): Ditto.
2189         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2190         * config/i386/i386-builtin.def: Update builtin definitions.
2191         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2192         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2193         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2194         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2195         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2196         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2197         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2198         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2199         * config/i386/subst.md:
2200         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2201         (round_saeonly_sd_mask_op4): Ditto.
2202         (round_saeonly_expand_operand5): Ditto.
2203         (round_saeonly_expand): Update.
2205 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2207         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2209 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2211         PR rtl-optimization/87871
2212         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2214 2018-11-05  Paul Koning  <ni1d@arrl.net>
2216         * doc/sourcebuild.texi (target attributes): Document new "inf"
2217         effective target keyword.
2219 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2221         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2222         moves involving the CC reg.
2224 2018-11-05  Richard Biener  <rguenther@suse.de>
2226         PR tree-optimization/87873
2227         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2228         argument.
2229         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2230         * tree-vect-loop.c (vect_transform_loop): When splitting the
2231         loop exit also create forwarder PHIs for constants.
2232         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2233         Handle constant to_arg, add extra checking we match up the correct
2234         PHIs.
2236 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2238         * config/s390/s390.md: QImode and HImode for load on condition.
2240 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2242         * config/s390/predicates.md: Fix typo.
2243         * config/s390/s390.md: Allow immediates for load on condition.
2245 2018-11-05  Martin Liska  <mliska@suse.cz>
2247         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2248         * bitmap.h (struct bitmap_usage): Likewise.
2249         * ggc-common.c (SCALE): Remove.
2250         (LABEL): Likewise.
2251         (struct ggc_usage): Use SIZE_AMOUNT. And update
2252         compare method.
2253         * ggc-page.c (SCALE): Remove.
2254         (STAT_LABEL): Remove.
2255         (ggc_print_statistics): Use SIZE_AMOUNT.
2256         * gimple.h (SCALE): Remove.
2257         (LABEL): Likewise.
2258         * input.c (ONE_K): Remove.
2259         (ONE_M): Likewise.
2260         (SCALE): Likewise.
2261         (STAT_LABEL): Likewise.
2262         (FORMAT_AMOUNT): Likewise.
2263         (dump_line_table_statistics): Use SIZE_AMOUNT.
2264         * mem-stats.h (struct mem_usage): Likewise.
2265         * rtl.c (dump_rtx_statistics): Likewise.
2266         (rtx_alloc_counts): Change type to size_t.
2267         (rtx_alloc_sizes): Likewise.
2268         (rtx_count_cmp): New.
2269         (dump_rtx_statistics): Sort first based on counts.
2270         * tree.c (tree_nodes_cmp): New.
2271         (tree_codes_cmp): New.
2272         (dump_tree_statistics): Sort first based on counts.
2273         * system.h (ONE_K): New.
2274         (ONE_M): Likewise.
2275         (SIZE_SCALE): Likewise.
2276         (SIZE_LABEL): Likewise.
2277         (SIZE_AMOUNT): Likewise.
2278         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2279         * tree-dfa.c (dump_dfa_stats): Likewise.
2280         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2281         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2282         * tree.c (dump_tree_statistics): Likewise.
2283         * vec.c (struct vec_usage): Likewise.
2284         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2285         trigger a -Werror=format-overflow with
2286         --enable-gather-detailed-stats.
2288 2018-11-05  Martin Liska  <mliska@suse.cz>
2290         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2291         Return T *.
2292         * vec.c (struct vec_usage): Register m_element_size.
2293         (vec_prefix::register_overhead): New arguments: elements and
2294         element_size.
2295         (vec_prefix::release_overhead): Subtract elements.
2296         * vec.h (struct vec_prefix): Change signature.
2297         (va_heap::reserve): Pass proper arguments.
2298         (va_heap::release): Likewise.
2300 2018-11-05  Martin Liska  <mliska@suse.cz>
2302         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2303         style.
2304         * vec.c: Likewise.
2306 2018-11-05  Richard Biener  <rguenther@suse.de>
2308         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2309         prototype.
2310         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2311         whether anything was done.
2312         (scev_const_prop): Remove constant propagation part, fold
2313         remains into ...
2314         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2315         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2316         conditionally.
2318 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2320         PR tree-optimization/87859
2321         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2322         only_constants and first_nonmergeable_order members.
2323         (merged_store_group::merged_store_group): Initialize them.
2324         (merged_store_group::do_merge): Clear only_constants member if
2325         adding something other than INTEGER_CST store.
2326         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2327         stores with order >= first_nonmergeable_order.  Use
2328         merged_store->only_constants instead of always recomputing it.
2329         Set merged_store->first_nonmergeable_order if we've skipped any
2330         stores.  Attempt to merge overlapping INTEGER_CST stores that
2331         we would otherwise skip.
2333         PR sanitizer/87837
2334         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2336 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2338         PR target/87853
2339         * config/i386/emmintrin.h (__v16qs): New to cope with option
2340         -funsigned-char.
2341         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2342         (_mm_cmplt_epi8): Likewise.
2343         (_mm_cmpgt_epi8): Likewise.
2345 2018-11-05  Richard Biener  <rguenther@suse.de>
2347         PR rtl-optimization/87852
2348         * fwprop.c (use_killed_between): Only consider single-defs of the
2349         use whose definition statement dominates the use.
2351 2018-11-05  Martin Liska  <mliska@suse.cz>
2353         PR web/87829
2354         * doc/invoke.texi: Remove options that are
2355         not disabled with -Os.
2357 2018-11-05  Martin Liska  <mliska@suse.cz>
2359         PR c/87811
2360         * doc/extend.texi: Update constrain about the last argument
2361         of __builtin_expect_with_probability.
2363 2018-11-05  Martin Liska  <mliska@suse.cz>
2365         PR c/87811
2366         * predict.c (expr_expected_value_1): Verify
2367         that last argument is a real constants and emit
2368         error.
2370 2018-11-05  Martin Liska  <mliska@suse.cz>
2372         PR gcov-profile/77698
2373         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2374         only in LTO mode.
2376 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2378         PR tree-optimization/86572
2379         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2381 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2383         PR tree-optimization/87672
2384         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2385         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2387 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2389         PR middle-end/58372
2390         * cfgexpand.c (pass_expand::execute): Move the call to
2391         finish_eh_generation in front of the call to expand_stack_alignment.
2393 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2395         * common/config/i386/i386-common.c (processor_alias_table): Add
2396         znver2 entry.
2397         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2398         (case ${target}): Add znver2.
2399         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2400         -march=native recognize znver2 processors.
2401         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2402         * config/i386/i386.c (m_znver2): New definition.
2403         (m_ZNVER): New definition.
2404         (m_AMD_MULTIPLE): Includes m_znver2.
2405         (processor_cost_table): Add znver2 entry.
2406         (processor_target_table): Add znver2 entry.
2407         (get_builtin_code_for_version): Set priority for
2408         PROCESSOR_ZNVER2.
2409         (processor_model): Add M_AMDFAM17H_ZNVER2.
2410         (arch_names_table): Ditto.
2411         (ix86_reassociation_width): Include znver2.
2412         * config/i386/i386.h (TARGET_znver2): New definition.
2413         (struct ix86_size_cost): Add TARGET_ZNVER2.
2414         (enum processor_type): Add PROCESSOR_ZNVER2.
2415         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2416         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2417         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2418         (ix86_adjust_cost): Add znver2.
2419         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2420         * gcc/doc/extend.texi: Add details about znver2.
2421         * gcc/doc/invoke.texi: Add details about znver2.
2423 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2425         PR target/87079
2427         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2428         pattern.
2430 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2432         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2433         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2435 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2437         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2438         shrink-wrapping when -mtrack-speculation.
2440 2018-11-02  Richard Biener  <rguenther@suse.de>
2442         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2443         (pop_cost_one_pair): Do not free pair.
2444         (pop_best_coalesce): Likewise.
2445         (create_coalesce_list): Initialize obstack.
2446         (delete_coalesce_list): Free obstack.
2447         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2448         (add_cost_one_coalesce): Likewise.
2449         (struct live_track): Remove bitmap pointer indirections.
2450         (new_live_track): Adjust.
2451         (delete_live_track): Likewise.
2452         (live_track_remove_partition): Likewise.
2453         (live_track_add_partition): Likewise.
2454         (live_track_live_p): Likewise.
2455         (live_track_process_def): Likewise.
2456         (live_track_clear_base_vars): Likewise.
2458 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2460         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2461         * configure: Regenerate.
2462         * config.in: Regenerate.
2463         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2464         HAVE_AS_REGISTER_PSEUDO_OP guard.
2465         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2466         !HAVE_AS_REGISTER_PSEUDO_OP support.
2467         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2468         guard.
2470 2018-11-02  Richard Biener  <rguenther@suse.de>
2472         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2473         with is_type_die.
2475 2018-11-02  Richard Biener  <rguenther@suse.de>
2477         PR tree-optimization/87776
2478         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2479         executable when iterating but running into rpo-vn-max-loop-depth
2480         and not eliding the iteration.
2482 2018-11-30  Jan Hubicka  <jh@suse.cz>
2484         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2485         TYPE_DECL.
2487 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2489         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2490         prototype.
2491         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2492         Combine with rs6000_address_for_fpconvert.
2493         (rs6000_address_for_fpconvert): Combine with
2494         rs6000_force_indexed_or_indirect_mem.
2495         (rs6000_expand_vector_init): Change function call from
2496         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2497         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2498         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2499         (floatsi<mode>2_lfiwax_mem): Ditto.
2500         (floatunssi<mode>2_lfiwzx): Ditto.
2501         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2502         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2503         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2504         (fix_trunc<mode>si2_stfiwx): Ditto.
2505         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2506         (float_<mode>si2_hw): Ditto.
2507         (floatuns_<mode>si2_hw): Ditto.
2508         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2509         (vsx_splat_<mode>): Ditto.
2511 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2513         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2514         at end of assembler input text.
2515         * configure: Regenerate.
2517 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2519         PR tree-optimization/87826
2520         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2521         negative or larger or equal to type's precision.
2523 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2525         * opts.c (default_options_table): Do not enable
2526         OPT_fdelayed_branch at -Og.
2527         * doc/invoke.texi (-fdelayed-branch): Document it.
2529 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2531         * optabs-libfuncs.c (build_libfunc_function_visibility):
2532         New, split out from...
2533         (build_libfunc_function): ... here.
2534         (init_one_libfunc_visibility): New, split out from ...
2535         (init_one_libfunc): ... here.
2537         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2538         scratch register need not be early-clobber.  Document the reason
2539         why we cannot use ST<OP>.
2541 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2543         PR bootstrap/82856
2544         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2545         line for second argument of AC_DEFINE_UNQUOTED.
2546         * doc/install.texi (Tools/packages necessary for modifying GCC):
2547         Update to autoconf 2.69 and automake 1.15.1.
2548         * aclocal.m4, config.in, configure: Regenerate.
2550 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2552         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2553         initialization.
2555 2018-10-31  Martin Liska  <mliska@suse.cz>
2557         PR driver/83193
2558         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2559         Add new argument invalid_extension.
2560         (aarch64_get_all_extension_candidates): New function.
2561         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2562         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2563         new argument.
2564         (aarch64_get_all_extension_candidates): New function.
2565         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2566         argument invalid_extension.
2567         (aarch64_parse_cpu): Likewise.
2568         (aarch64_print_hint_for_extensions): New function.
2569         (aarch64_validate_mcpu): Provide hint about invalid extension.
2570         (aarch64_validate_march): Likewise.
2571         (aarch64_handle_attr_arch): Pass new argument.
2572         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2573         (aarch64_handle_attr_isa_flags): Likewise.
2575 2018-10-31  Richard Biener  <rguenther@suse.de>
2577         PR middle-end/70359
2578         PR middle-end/86270
2579         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2580         copy generation to useful cases.  Place the copy before
2581         the definition of the backedge value when possible.
2583 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2585         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2586         * config/arc/arc.c (arc_active_insn): New function.
2587         (check_store_cacheline_hazard): Likewise.
2588         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2589         (arc_override_options): Disable delay slot scheduler for older
2590         A7.
2591         (arc_store_addr_hazard_p): New implementation, old one renamed to
2592         ...
2593         (arc_store_addr_hazard_internal_p): Renamed.
2594         (arc_reorg): Don't combine into brcc instructions which are part
2595         of hardware hazard solution.
2596         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2597         (tune_arc700): Likewise.
2598         * config/arc/arc.opt (arc7xx): New tune value.
2599         * config/arc/arc700.md: Improve A7 scheduler.
2601 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2603         * config/arc/arc.c (arc_override_options): Remove
2604         TARGET_COMPACT_CASESI.
2605         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2606         (CASE_VECTOR_MODE): Likewise.
2607         (CASE_VECTOR_PC_RELATIVE): Likewise.
2608         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2609         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2610         (ADDR_VEC_ALIGN): Update.
2611         (ASM_OUTPUT_CASE_LABEL): Undefine.
2612         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2613         (TARGET_BI_BIH): Define.
2614         (DEFAULT_BRANCH_INDEX): Likewise.
2615         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2616         instructions, remove compact_casesi use case.
2617         (casesi_compact_jump): Remove.
2618         (casesi_dispatch): New pattern.
2619         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2620         compact_casesi option.
2621         * doc/invoke.texi: Document mbranch-index option.
2623 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2625         * config/arc/arc.c (arc_get_tp): Remove function.
2626         (arc_emit_call_tls_get_addr): Likewise.
2627         (arc_call_tls_get_addr): New function.
2628         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2629         * config/arc/arc.md (tls_load_tp_soft): Remove.
2630         (tls_gd_get_addr): Likewise.
2632 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2634         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2635         (mulsi3_600_lib): Remove pattern.
2636         (umulsi3_highpart_600_lib_le): Likewise.
2637         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2638         (umulsidi3): Remove call to umulsidi3_600_lib.
2639         (umulsidi3_600_lib): Remove pattern.
2640         (peephole2): Remove peephole using the above deprecated patterns.
2642 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2644         PR target/87374
2645         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2646         use of -mslow-flash-data and -mword-relocations.
2647         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2648         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2649         flag_pic.
2650         * doc/invoke.texi (-mword-relocations): Mention conflict with
2651         -mslow-flash-data.
2652         (-mslow-flash-data): Reciprocally.
2654 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2656         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2657         16-byte modes held in GP registers to use an even regno.
2659         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2660         (aarch64_atomic_ldop_supported_p): Remove.
2661         (aarch64_gen_atomic_ldop): Remove.
2662         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2663         Fully expand LSE operations here.
2664         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2665         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2666         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2667         and use ATOMIC_LDOP instead; use register_operand for the input;
2668         drop the split and emit insns directly.
2669         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2670         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2671         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2673         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2674         (aarch64_gen_atomic_ldop): Don't call it.
2675         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2676         Use aarch64_reg_or_zero.
2677         (aarch64_atomic_exchange<ALLI>): Likewise.
2678         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2679         operand 0; use aarch64_reg_or_zero for input; merge ...
2680         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2682         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2683         (aarch64_split_compare_and_swap): Use it.
2684         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2685         test oldval against the proper predicate.
2686         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2687         Use nonmemory_operand for expected.
2688         (cas_short_expected_pred): New.
2689         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2690         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2691         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2692         (aarch64_plushi_operand): New.
2694         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2695         Force oldval into the rval register for TARGET_LSE; emit the compare
2696         during initial expansion so that it may be deleted if unused.
2697         (aarch64_gen_atomic_cas): Remove.
2698         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2699         Change =&r to +r for operand 0; use match_dup for operand 2;
2700         remove is_weak and mod_f operands as unused.  Drop the split
2701         and merge with...
2702         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2703         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2704         (@aarch64_atomic_cas<GPI>): Similarly.
2706 2018-10-31  Richard Biener  <rguenther@suse.de>
2708         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2709         using ABSU_EXPR.
2711 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2713         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2714         * config/aarch64/aarch64.md: Include saphira.md
2715         * config/aarch64/saphira.md: New file for pipeline description.
2717 2018-10-30  Martin Sebor  <msebor@redhat.com>
2719         PR middle-end/87041
2720         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2721         inlining context.
2722         (sprintf_dom_walker::compute_format_length):
2723         Avoid setting POSUNDER4K here.
2724         (get_destination_size): Handle null argument values.
2725         (get_user_idx_format): New function.
2726         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2727         functions, including user-defined with attribute format printf.
2728         Use %G to include inlining context.
2729         Set POSUNDER4K here.
2731 2018-10-30  Jan Hubicka  <jh@suse.cz>
2733         * params.def (lto-partitions): Bump from 32 to 128.
2735 2018-10-30  Jan Hubicka  <jh@suse.cz>
2737         * tree.c
2738         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2739         head in file.
2740         (free_lang_data_in_type): Forward declare.
2741         (fld_type_variant_equal_p): New function.
2742         (fld_type_variant): New function
2743         (fld_incomplete_types): New hash.
2744         (fld_incomplete_type_of): New function
2745         (fld_simplfied-type): New function.
2746         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2747         (free_lang_data): Allocate and free fld_incomplete_type; update call
2748         of free_lang_data_in_decl.
2750 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2752         * gcov.c (output_lines): Remove duplicate line.
2754 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2756         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2757         if not in indexed or indirect form.
2758         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2759         (bswapdi2_store): Ditto.
2760         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2761         helper function.
2762         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2763         Prototype for helper function.
2765 2018-10-30  Martin Sebor  <msebor@redhat.com>
2767         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2768         (target, pragma GCC optimize, pragma GCC target): Ditto.
2770 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2772         * doc/extend.texi: Fix prototype and description of
2773         __builtin_expect_with_probability.
2775 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2777         * cgraph.h (clone_function_name_1): Replaced by new
2778           clone_function_name_numbered that takes name as string; for
2779           privatize_symbol_name_1 use only.
2780           (clone_function_name): Renamed to
2781           clone_function_name_numbered to be explicit about numbering.
2782           (clone_function_name): New two-argument function that does
2783           not number its output.
2784           (clone_function_name): New three-argument function that
2785           takes a number to append to its output.
2786         * cgraphclones.c (duplicate_thunk_for_node):
2787           (clone_function_name_1): Renamed.
2788           (clone_function_name_numbered): Two new functions.
2789           (clone_function_name): Improved documentation.
2790           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2791         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2792         * final.c (final_scan_insn_1): Use the new clone_function_name
2793           without numbering.
2794         * multiple_target.c (create_dispatcher_calls): Ditto.
2795           (create_target_clone): Ditto.
2796         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2797         * omp-low.c (create_omp_child_function_name): Ditto.
2798         * omp-simd-clone.c (simd_clone_create): Ditto.
2799         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2800           new clone_function_name without numbering.
2802 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2804         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2805         Assert that the allocation size is not zero.
2807 2018-10-30  Richard Biener  <rguenther@suse.de>
2809         PR tree-optimization/87800
2810         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2811         non-induction or reduction PHIs.
2813 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2815         * config/aarch64/falkor-tag-collision-avoidance.c
2816         (execute_tag_collision_avoidance): Call df_note_add_problem.
2818 2018-10-30  Martin Liska  <mliska@suse.cz>
2820         * doc/extend.texi: Fix typo in documentation
2821         of __builtin_expect_with_probability.
2823 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2825         PR c++/87721
2826         * input.c (get_substring_ranges_for_loc): Detect if
2827         linemap_resolve_location gives us a NULL map, and reject
2828         this case.
2830 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2832         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2834 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2835             Martin Sebor  <msebor@redhat.com>
2836             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2838         * doc/ux.texi (Quoting): New subsection, adapted from material at
2839         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2840         MartinSebor and ManuelLopezIbanez.
2841         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2842         translation.
2844 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2846         PR middle-end/87469
2847         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2848         max value.
2850 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2852         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2854 2018-10-29  Paul Koning  <ni1d@arrl.net>
2856         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2858 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2860         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2861         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2862         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2863         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2864         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2865         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2866         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2867         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2868         Change 'vector' to '__vector'.
2869         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2870         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2871         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2872         _mm_avg_pu16): Likewise.  And, whitespace corrections.
2874 2018-10-29  Richard Biener  <rguenther@suse.de>
2876         PR tree-optimization/87785
2877         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2878         internal defs.
2880 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2882         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2884 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2886         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2887         ports configuration macro, defaults to "gnu".
2888         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2889         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2891 2018-10-29  Olivier Hainque  <hainque@adacore.com>
2893         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2895 2018-10-29  Richard Biener  <rguenther@suse.de>
2897         PR tree-optimization/87790
2898         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2899         (vect_make_slp_decision): Adjust.
2900         (vect_slp_analyze_bb_1): Likewise.
2901         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2902         edges.
2904 2018-10-29  Richard Biener  <rguenther@suse.de>
2906         PR tree-optimization/87785
2907         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2908         and processing.
2909         (vect_build_slp_tree): Likewise.
2910         (vect_gather_slp_loads): New function.
2911         (vect_analyze_slp_instance): Gather loads separately from the
2912         SLP tree build.
2914 2018-10-29  Martin Liska  <mliska@suse.cz>
2916         * Makefile.in: Make dependency to json.o.
2917         * doc/gcov.texi: Document new JSON format, remove
2918         old intermediate format documentation.
2919         * gcov.c (struct function_info): Come up with m_name and
2920         m_demangled_name.
2921         (function_info::function_info): Initialize it.
2922         (function_info::~function_info): Release it.
2923         (main): Rename flag_intermediate_format to flag_json_format.
2924         (print_usage): Describe --json-format.
2925         (process_args): Set flag_json_format.
2926         (output_intermediate_line): Remove.
2927         (output_intermediate_json_line): Likewise.
2928         (get_gcov_intermediate_filename): Return new extension
2929         ".gcov.json.gz".
2930         (output_intermediate_file): Implement JSON emission.
2931         (output_json_intermediate_file): Implement JSON emission.
2932         (generate_results): Use ::get_name for function name.
2933         Handle JSON output file.
2934         (read_graph_file): Use ::get_name instead of cplus_demangle.
2935         (read_count_file): Likewise.
2936         (solve_flow_graph): Likewise.
2937         (add_line_counts): Likewise.
2938         (accumulate_line_counts): Use new flag_json_format.
2939         (output_function_details): Use ::get_name instead of cplus_demangle.
2940         (output_lines): Likewise.
2941         * json.cc (test_writing_literals): Add new tests.
2942         * json.h (class literal): Add new boolean constructor.
2944 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
2946         PR rtl-optimization/87701
2947         PR rtl-optimization/87780
2948         * combine.c (make_more_copies): Rewrite.
2950 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
2952         * doc/generic.texi (ABSU_EXPR): Document.
2953         * match.pd (absu(x)*absu(x) -> x*x): Handle.
2954         (absu(absu(X)) -> absu(X)): Likewise.
2955         (absu(-X) -> absu(X)): Likewise.
2956         (absu(X)  where X is nonnegative -> X): Likewise.
2958 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
2960         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2961         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2962         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2963         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2964         (s-tm-texi): Also check timestamp on d-target.def.
2965         (generated_files): Add TM_D_H and d-target-hooks-def.h.
2966         (build/genhooks.o): Also depend on D_TARGET_DEF.
2967         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2968         variables.
2969         * config/aarch64/aarch64-d.c: New file.
2970         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2971         Define.
2972         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2973         prototype.
2974         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2975         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2976         * config/arm/arm-d.c: New file.
2977         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2978         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2979         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2980         * config/arm/t-arm (arm-d.o): New rule.
2981         * config/default-d.c: New file.
2982         * config/glibc-d.c: New file.
2983         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2984         * config/i386/i386-d.c: New file.
2985         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2986         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2987         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2988         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2989         * config/i386/t-i386 (i386-d.o): New rule.
2990         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2991         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2992         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2993         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2994         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2995         * config/mips/mips-d.c: New file.
2996         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2997         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2998         * config/mips/t-mips (mips-d.o): New rule.
2999         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3000         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3001         * config/powerpcspe/powerpcspe-d.c: New file.
3002         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3003         New prototype.
3004         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3005         Support GNU D by using 0 as the language type.
3006         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3007         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3008         * config/riscv/riscv-d.c: New file.
3009         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3010         prototype.
3011         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3012         * config/riscv/t-riscv (riscv-d.o): New rule.
3013         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3014         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3015         * config/rs6000/rs6000-d.c: New file.
3016         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3017         prototype.
3018         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3019         Support GNU D by using 0 as the language type.
3020         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3021         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3022         * config/s390/s390-d.c: New file.
3023         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3024         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3025         * config/s390/t-s390 (s390-d.o): New rule.
3026         * config/sparc/sparc-d.c: New file.
3027         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3028         prototype.
3029         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3030         * config/sparc/t-sparc (sparc-d.o): New rule.
3031         * config/t-glibc (glibc-d.o): New rule.
3032         * configure: Regenerated.
3033         * configure.ac (tm_d_file): New variable.
3034         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3035         * doc/contrib.texi (Contributors): Add self for the D frontend.
3036         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3037         * doc/install.texi (Configuration): Mention libphobos as an option for
3038         --enable-shared.  Mention d as an option for --enable-languages.
3039         (Testing): Mention check-d as a target.
3040         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3041         name suffixes.  Mention d as a -x option.
3042         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3043         * doc/standards.texi (Standards): Add section on D language.
3044         * doc/tm.texi: Regenerated.
3045         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3046         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3047         * dwarf2out.c (is_dlang): New function.
3048         (gen_compile_unit_die): Use DW_LANG_D for D.
3049         (declare_in_namespace): Return module die for D, instead of adding
3050         extra declarations into the namespace.
3051         (gen_namespace_die): Generate DW_TAG_module for D.
3052         (gen_decl_die): Handle CONST_DECLSs for D.
3053         (dwarf2out_decl): Likewise.
3054         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3055         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3056         kinds of aggregates.
3057         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3058         * genhooks.c: Include d/d-target.def.
3060 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3062         PR target/85669
3063         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3064         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3065         (STACK_DYNAMIC_OFFSET): Likewise.
3067 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3069         PR target/80024
3070         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3071         error message.
3073 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3075         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3076         return type and other typos.
3078 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3080         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3081         a shorter sequence with fewer branches.
3082         (emit_final_str_compare_gpr): Ditto.
3084 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3086         * config/rs6000/tmmintrin.h: New file.
3087         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3089 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3091         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3092         * config/rs6000/xmmintrin.h: Likewise.
3094 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3096         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3098 2018-10-26  Richard Biener  <rguenther@suse.de>
3100         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3101         and wrapper.
3102         (vect_mark_slp_stmts_relevant): Likewise.
3103         (vect_detect_hybrid_slp_stmts): Likewise.
3104         (vect_bb_slp_scalar_cost): Likewise.
3105         (vect_remove_slp_scalar_calls): Likewise.
3107 2018-10-26  Jan Hubicka  <jh@suse.cz>
3109         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3110         (warn_types_mismatch): Fix walk of DECL_NAME.
3111         (odr_types_equivalent_p): Fix overactive assert.
3113 2018-10-26  Richard Biener  <rguenther@suse.de>
3115         PR tree-optimization/87105
3116         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3117         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3118         refcnt.
3119         (vect_create_new_slp_node): Initialize refcnt to one.
3120         (bst_traits): Move.
3121         (scalar_stmts_set_t, bst_fail): Remove.
3122         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3123         (vect_build_slp_tree): Add bst_map argument and lookup
3124         already created SLP nodes.
3125         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3126         addresses.
3127         (vect_slp_rearrange_stmts): Handle a SLP graph.
3128         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3129         the CSE map.  Fix indenting.
3130         (vect_schedule_slp_instance): Add short-cut.
3132 2018-10-26  Martin Liska  <mliska@suse.cz>
3134         PR testsuite/86158
3135         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3136         addr_expr and not with pointers.
3138 2018-10-26  Jan Hubicka  <jh@suse.cz>
3140         * tree.c (free_lang_data_in_type): Only check main variants.
3141         * ipa-devirt.c (warn_odr): Make static.
3142         (types_same_for_odr): Drop strict variant.
3143         (types_odr_comparable): Likewise.
3144         (odr_or_derived_type_p): Look for main variants.
3145         (odr_name_hasher::equal): Cleanup comment.
3146         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3147         variants.
3148         (type_variants_equivalent_p): break out from ...
3149         (odr_types_equivalent): ... here; go for main variants where needed.
3150         (warn_odr): ... here; turn static.
3151         (warn_types_mismatch): Compare mangled names of main variants.
3152         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3153         (type_with_linkage_p): Sanity check that we look at main variant.
3154         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3155         * tree.h (types_same_for_odr): Drop strict argument.
3157 2018-10-26  Richard Biener  <rguenther@suse.de>
3159         PR tree-optimization/87746
3160         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3161         Simplify and fix WRT strided store groups with size not
3162         equal to step in element count.
3163         (vect_analyze_group_access_1): Dump the whole group.
3165 2018-10-25  Carl Love  <cel@us.ibm.com>
3167         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3168         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3169         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3170         precicion entry for each overloaded builtin.
3171         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3172         VSCEDPUO): Rename overloaded name.
3173         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3174         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3175         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3176         define_expand for xscmpexqp instruction.
3177         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3179 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3180             Jinsong Ji <jji@us.ibm.com>
3182         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3183         function with vec_sl.
3184         (_mm_slli_epi32): Likewise.
3185         (_mm_slli_epi64): Likewise.
3186         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3187         (_mm_srai_epi32): Likewise.
3188         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3189         (_mm_srli_epi32): Likewise.
3190         (_mm_srli_epi64): Likewise.
3191         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3192         (_mm_sll_epi32): Likewise.
3193         (_mm_sll_epi64): Likewise.
3194         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3195         (_mm_sra_epi32): Likewise.
3196         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3197         (_mm_srl_epi32): Likewise.
3198         (_mm_srl_epi64): Likewise.
3200 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3201             Jinsong Ji <jji@us.ibm.com>
3203         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3204         comparison operators with vec_cmp* for compatibility due to
3205         unfortunate history; clean up formatting and use types more
3206         appropriately.
3207         (_mm_sll_epi32): Likewise.
3208         (_mm_sll_epi64): Likewise.
3209         (_mm_srl_epi16): Likewise.
3210         (_mm_srl_epi32): Likewise.
3211         (_mm_srl_epi64): Likewise.
3213 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3214             Jinsong Ji <jji@us.ibm.com>
3216         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3217         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3218         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3219         function vec_vcmpgtfp.
3220         (_mm_max_ps): Likewise.
3222 2018-10-25  Jeff Law  <law@redhat.com>
3224         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3225         if returning false.
3227 2018-10-25  Martin Sebor  <msebor@redhat.com>
3229         * doc/extend.texi (aligned): Expand attribute description.
3230         (Alignment): Rename section.  Discuss function arguments.
3232 2018-10-25  Jan Hubicka  <jh@suse.cz>
3234         * ipa-devirt.c (main_odr_variant): Remove.
3235         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3236         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3237         Drop use of main_odr_variant.
3238         (add_type_duplicate): Silence confused warnings on integer types.
3239         (get_odr_type): Always look for main variant.
3240         (register_odr_type): Simplify.
3242 2018-10-25  Richard Biener  <rguenther@suse.de>
3244         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3245         Initialize ng to silence error with release checking bootstrap.
3247 2018-10-25  Richard Biener  <rguenther@suse.de>
3249         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3250         (tree_if_conversion): Run CSE on the if-converted loop body.
3252 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3254         * config/s390/constraints.md (ZL): New constraint.
3255         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3256         operands.
3257         * config/s390/s390.md (movdi_larl): Remove.
3258         (movdi_64): Add the LARL alternative.
3260 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3262         PR bootstrap/87747
3263         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3264         (RTX_CODE_HWINT_P): New macro.
3265         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3267 2018-10-25  Jan Hubicka  <jh@suse.cz>
3269         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3270         is anonymous.
3272 2018-10-25  Richard Biener  <rguenther@suse.de>
3274         PR tree-optimization/87665
3275         PR tree-optimization/87745
3276         * tree-vectorizer.h (get_earlier_stmt): Remove.
3277         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3279 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3281         * options.texi (Deprecated): Move list to Var section.
3283 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3284             Jinsong Ji <jji@us.ibm.com>
3286         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3287         __vector long to __vector long long.
3288         (_mm_cvtpd_ps): Likewise.
3289         (_mm_cvttpd_epi32): Likewise.
3290         (_mm_cvtpi32_pd): Likewise.
3291         (_mm_unpackhi_epi64): Likewise.
3292         (_mm_unpacklo_epi64): Likewise.
3294 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3296         PR rtl-optimization/87720
3297         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3299 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3301         * gimple-ssa-isolate-paths.c
3302         (find_implicit_erroneous_behavior): Do not change code if the
3303         pass is running for warnings only.
3304         (find_explicit_erroneous_behavior): Likewise.
3306 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3308         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3309         Define as rs6000_mangle_decl_assembler_name.
3310         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3311         long double to IEEE long double, switch the names of the long
3312         double built-in functions to be <func>f128 instead of <func>l.
3314 2018-10-24  Martin Sebor  <msebor@redhat.com>
3316         * doc/extend.texi (nonnull): List no-argument form.  Reference
3317         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3319 2018-10-24  Richard Biener  <rguenther@suse.de>
3321         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3323 2018-10-24  Martin Liska  <mliska@suse.cz>
3325         PR tree-optimization/84436
3326         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3327         Remove.
3328         (switch_conversion::contains_linear_function_p): New.
3329         (switch_conversion::build_one_array): Support linear
3330         transformation on input.
3331         * tree-switch-conversion.h (struct switch_conversion): Add
3332         contains_linear_function_p declaration.
3334 2018-10-24  Richard Biener  <rguenther@suse.de>
3336         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3337         if its argument is CONSTANT_CLASS_P.
3339 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3341         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3342         it is wrong for forward declarations.
3344 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3346         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3347         SYMBOL_REF_P () check.
3349 2018-10-24  Richard Biener  <rguenther@suse.de>
3351         PR tree-optimization/87105
3352         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3353         dump classification.
3354         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3355         stores by splitting the affected group after the fact.
3356         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3357         fail the SLP build because of size constraints.
3359 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3361         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3362         * configure: Regenerate.
3363         * config.in: Regenerate.
3364         * varasm.c (mergeable_string_section): Use readonly_data_section
3365         if linker doesn't support SHF_MERGE with alignment > 8.
3366         (mergeable_constant_section): Likewise.
3368 2018-10-24  Richard Biener  <rguenther@suse.de>
3370         PR tree-optimization/84013
3371         * tree-ssa-structalias.c (struct msdi_data): New struct for
3372         marshalling data to walk_stmt_load_store_ops.
3373         (maybe_set_dependence_info): Refactor as callback for
3374         walk_stmt_load_store_ops.
3375         (compute_dependence_clique): Set restrict info on all stmt kinds.
3377 2018-10-24  Martin Liska  <mliska@suse.cz>
3379         * cgraph.c (cgraph_node::dump):
3380         Remove reduntant dumps and make tp_first_run dump more compact.
3382 2018-10-24  Richard Biener  <rguenther@suse.de>
3384         PR tree-optimization/87665
3385         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3386         to reflect reality.
3388 2018-10-12  Jeff Law  <law@redhat.com>
3390         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3391         for H8/S.
3393 2018-10-23  Richard Biener  <rguenther@suse.de>
3395         * tree-vrp.c (add_assert_info): Guard dump_printf with
3396         dump_enabled_p.
3397         * gimple-ssa-evrp-analyze.c
3398         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3399         Use value_range::ignore_equivs_equal_p.
3401 2018-10-23  Richard Biener  <rguenther@suse.de>
3403         PR tree-optimization/87105
3404         PR tree-optimization/87608
3405         * passes.def (pass_all_early_optimizations): Add early phi-opt
3406         after dce.
3407         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3408         addition to debug stmts.
3409         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3410         and abs replacement early.
3411         * tree-cfg.c (gimple_empty_block_p): Likewise.
3413 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3415         PR target/86383
3416         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3417         specified to configure.
3418         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3420 2018-10-23  Richard Biener  <rguenther@suse.de>
3422         PR tree-optimization/87700
3423         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3425 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3427         PR target/87674
3428         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3429         second argument from __mmask16 to __mmask8.
3430         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3431         _mm_mask_packs_epi32): Likewise.
3432         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3433         Likewise.
3434         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3436 2018-10-23  Richard Biener  <rguenther@suse.de>
3438         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3440 2018-10-23  Richard Biener  <rguenther@suse.de>
3442         PR tree-optimization/86144
3443         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3444         over simd attribute.
3446 2018-10-23  Richard Biener  <rguenther@suse.de>
3448         PR tree-optimization/87693
3449         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3450         the case we do not find the taken edge.
3452 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3453             Jinsong Ji  <jji@us.ibm.com>
3455         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3456         (_mm_store_pd): Use unaligned vector type for pointer cast.
3457         (_mm_maskmoveu_si128): Likewise.
3458         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3459         (_mm_store_ps): Use unaligned vector type for pointer cast.
3461 2018-10-22  Paul Koning  <ni1d@arrl.net>
3463         * symtab.c (symtab_node::increase_alignment): Correct max
3464         alignment check.
3466 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3468         PR tree-optimization/87633
3469         * match.pd: Do not generate unordered integer comparisons.
3471 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3473         PR rtl-optimization/87600
3474         * combine.c: Add include of expr.h.
3475         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3476         register to a pseudo.
3477         (make_more_copies): New function, add a copy to a new pseudo after
3478         the moves from hard registers into pseudos.
3479         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3480         later.  Call make_more_copies.
3482 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3484         * lra-constraints.c (process_alt_operands): New local array,
3485         matching_early_clobber.  Check matching_early_clobber before
3486         decrementing reject, and set matching_early_clobber after.
3488 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3490         PR target/87598
3491         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3492         call output_addr_const and hope for the best.
3494 2018-10-22  Richard Biener  <rguenther@suse.de>
3496         * gimple-ssa-evrp-analyze.c
3497         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3498         smarter about what ranges to use.
3499         * tree-vrp.c (add_assert_info): Dump here.
3500         (register_edge_assert_for_2): Instead of here at multiple but
3501         not all places.
3503         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3504         * gcc.dg/predict-6.c: Adjust.
3505         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3506         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3507         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3509 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3510         Richard Biener  <rguenther@suse.de>
3512         * bitmap.h: Update data structure documentation, including a
3513         description of bitmap views as either linked-lists or splay trees.
3514         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3515         (struct bitmap_head_def): Likewise.
3516         (bitmap_list_view, bitmap_tree_view): New prototypes.
3517         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3518         tree_form fields.
3519         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3520         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3521         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3522         released bitmap element here.
3523         (bitmap_element_free): Remove.
3524         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3525         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3526         this function similar ones such that linked-list bitmap implementation
3527         functions are grouped.
3528         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3529         and moved for grouping.
3530         (bitmap_list_insert_element_after): Renamed from
3531         bitmap_elt_insert_after, and moved for grouping.
3532         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3533         (bitmap_tree_link_left, bitmap_tree_link_right,
3534         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3535         bitmap_tree_link_element, bitmap_tree_unlink_element,
3536         bitmap_tree_find_element): New functions for splay-tree bitmap
3537         implementation.
3538         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3539         Renamed and moved, see above entries.
3540         (bitmap_tree_listify_from): New function to convert part of a splay
3541         tree bitmap to a linked-list bitmap.
3542         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3543         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3544         (bitmap_find_bit): Remove.
3545         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3546         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3547         Handle splay tree bitmaps.
3548         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3549         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3550         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3551         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3552         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3553         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3554         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3555         corresponding changes to use linked-list specific bitmap_element
3556         manipulation functions as applicable for efficiency.
3557         (bitmap_tree_to_vec): New function.
3558         (debug_bitmap_elt_file): New function split out from ...
3559         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3560         (bitmap_print): Likewise.
3562         PR tree-optimization/63155
3563         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3564         SSA edge worklists.
3565         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3566         in tree-view.
3568 2018-10-22  Martin Liska  <mliska@suse.cz>
3570         PR tree-optimization/87686
3571         Revert
3572         2018-08-29  Martin Liska  <mliska@suse.cz>
3574         * tree-switch-conversion.c (switch_conversion::expand):
3575         Strenghten assumption about gswitch statements.
3577 2018-10-22  Martin Liska  <mliska@suse.cz>
3579         * ipa-icf.c (sem_item::compare_attributes): Remove.
3580         (sem_item::compare_referenced_symbol_properties): Use
3581         attribute_list_equal instead.
3582         (sem_function::equals_wpa): Likewise.
3583         * ipa-icf.h: Remove compare_attributes.
3585 2018-10-22  Richard Biener  <rguenther@suse.de>
3587         PR middle-end/87682
3588         * mem-stats.h (mem_usage::operator==): Fix pasto.
3590 2018-10-22  Richard Biener  <rguenther@suse.de>
3592         PR tree-optimization/87640
3593         * tree-vrp.c (set_value_range_with_overflow): Decompose
3594         incomplete result.
3595         (extract_range_from_binary_expr_1): Adjust.
3597 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3599         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3600         (stmt_can_throw_external): Likewise.
3601         (stmt_can_throw_internal): Likewise.
3602         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3603         (lower_eh_constructs_2): Likewise.
3604         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3605         (stmt_can_throw_external): Likewise.
3606         (stmt_can_throw_internal): Likewise.
3607         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3608         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3609         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3610         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3611         (pass_lower_eh_dispatch::execute): Pass cfun to
3612         stmt_can_throw_external.
3613         (cleanup_empty_eh): Likewise.
3614         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3615         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3616         stmt_can_throw_external instead of pushing it to cfun.
3617         (symbol_table::create_edge): Likewise.
3618         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3619         stmt_can_throw_internal.
3620         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3621         to stmt_could_throw_p.
3622         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3623         stmt_can_throw_internal.
3624         (pass_store_merging::execute): Likewise.
3625         * gimple-ssa-strength-reduction.c
3626         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3627         stmt_could_throw_p.
3628         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3629         stmt_can_throw_internal.
3630         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3631         to stmt_can_throw_external.
3632         (check_stmt): Pass cfun to stmt_could_throw_p.
3633         (check_stmt): Pass cfun to stmt_can_throw_external.
3634         (pass_nothrow::execute): Likewise.
3635         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3636         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3637         stmt_can_throw_internal.
3638         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3639         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3640         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3641         * tree-complex.c (expand_complex_libcall): Pass cfun to
3642         stmt_could_throw_p and to stmt_can_throw_internal.
3643         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3644         * tree-inline.c (copy_edges_for_bb): Likewise.
3645         (maybe_move_debug_stmts_to_successors): Likewise.
3646         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3647         stmt_could_throw_p.
3648         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3649         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3650         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3651         stmt_can_throw_internal.
3652         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3653         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3654         stmt_could_throw_p.
3655         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3656         stmt_can_throw_internal.
3657         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3658         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3659         stmt_could_throw_p.
3660         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3661         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3662         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3663         (convert_mult_to_fma_1): Likewise.
3664         (convert_to_divmod): Likewise.
3665         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3666         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3667         * tree-ssa-propagate.c
3668         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3669         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3670         (maybe_optimize_range_tests): Likewise.
3671         (linearize_expr_tree): Likewise.
3672         (reassociate_bb): Likewise.
3673         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3674         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3675         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3676         (handle_char_store): Likewise.
3677         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3678         stmt_can_throw_internal.
3679         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3680         stmt_could_throw_p.
3681         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3682         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3683         (vectorizable_simd_clone_call): Likewise.
3684         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3685         (gimple_stringop_fixed_value): Likewise.
3687 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3689         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3690         literal pool references.
3691         (s390_check_qrst_address): Adapt to the new behavior of
3692         s390_loadrelative_operand_p ().
3694 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3696         PR target/72782
3697         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3699 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3701         PR target/72782
3702         * config/i386/sse.md (*<code><mode>3_bcst): New.
3704 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3706         PR target/72782
3707         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3708         V4DI, V16SI and V8DI.
3709         (*sub<mode>3<mask_name>_bcst): New.
3710         (*add<mode>3<mask_name>_bcst): Likewise.
3712 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3713             Jinsong Ji  <jji@us.ibm.com>
3715         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3716         __m64 with __vector unsigned long long for compatibility.
3717         (_mm_movemask_epi8): Likewise.
3718         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3719         (_mm_cvttps_pi32): Likewise.
3720         (_mm_cvtpi32_ps): Likewise.
3721         (_mm_cvtps_pi16): Likewise.
3722         (_mm_loadh_pi): Likewise.
3723         (_mm_storeh_pi): Likewise.
3724         (_mm_movehl_ps): Likewise.
3725         (_mm_movelh_ps): Likewise.
3726         (_mm_loadl_pi): Likewise.
3727         (_mm_storel_pi): Likewise.
3728         (_mm_movemask_ps): Likewise.
3729         (_mm_shuffle_pi16): Likewise.
3731 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3733         PR target/72782
3734         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3735         __builtin_ia32_vfnmsubpd512_mask.
3736         (_mm512_mask_fnmsub_round_pd): Likewise.
3737         (_mm512_fnmsub_pd): Likewise.
3738         (_mm512_mask_fnmsub_pd): Likewise.
3739         (_mm512_maskz_fnmsub_round_pd): Use
3740         __builtin_ia32_vfnmsubpd512_maskz.
3741         (_mm512_maskz_fnmsub_pd): Likewise.
3742         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3743         (_mm512_mask_fnmsub_round_ps): Likewise.
3744         (_mm512_fnmsub_ps): Likewise.
3745         (_mm512_mask_fnmsub_ps): Likewise.
3746         (_mm512_maskz_fnmsub_round_ps): Use
3747         __builtin_ia32_vfnmsubps512_maskz.
3748         (_mm512_maskz_fnmsub_ps): Likewise.
3749         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3750         __builtin_ia32_vfnmsubpd256_mask.
3751         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3752         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3753         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3754         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3755         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3756         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3757         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3758         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3759         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3760         __builtin_ia32_vfnmsubpd.
3761         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3762         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3763         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3764         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3765         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3766         * config/i386/i386-builtin.def: Add
3767         __builtin_ia32_vfnmsubpd256_mask,
3768         __builtin_ia32_vfnmsubpd256_maskz,
3769         __builtin_ia32_vfnmsubpd128_mask,
3770         __builtin_ia32_vfnmsubpd128_maskz,
3771         __builtin_ia32_vfnmsubps256_mask,
3772         __builtin_ia32_vfnmsubps256_maskz,
3773         __builtin_ia32_vfnmsubps128_mask,
3774         __builtin_ia32_vfnmsubps128_maskz,
3775         __builtin_ia32_vfnmsubpd512_mask,
3776         __builtin_ia32_vfnmsubpd512_maskz,
3777         __builtin_ia32_vfnmsubps512_mask,
3778         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3779         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3780         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3781         __builtin_ia32_vfnmsubpd256.
3782         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3783         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3784         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3785         Likewise.
3786         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3787         Likewise.
3788         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3789         Likewise.
3790         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3792 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3794         PR target/72782
3795         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3796         __builtin_ia32_vfnmaddpd512_mask.
3797         (_mm512_mask_fnmadd_round_pd): Likewise.
3798         (_mm512_fnmadd_pd): Likewise.
3799         (_mm512_mask_fnmadd_pd): Likewise.
3800         (_mm512_maskz_fnmadd_round_pd): Use
3801         __builtin_ia32_vfnmaddpd512_maskz.
3802         (_mm512_maskz_fnmadd_pd): Likewise.
3803         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3804         (_mm512_mask_fnmadd_round_ps): Likewise.
3805         (_mm512_fnmadd_ps): Likewise.
3806         (_mm512_mask_fnmadd_ps): Likewise.
3807         (_mm512_maskz_fnmadd_round_ps): Use
3808         __builtin_ia32_vfnmaddps512_maskz.
3809         (_mm512_maskz_fnmadd_ps): Likewise.
3810         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3811         __builtin_ia32_vfnmaddpd256_mask.
3812         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3813         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3814         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3815         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3816         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3817         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3818         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3819         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3820         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3821         __builtin_ia32_vfnmaddpd.
3822         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3823         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3824         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3825         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3826         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3827         * config/i386/i386-builtin.def: Add
3828         __builtin_ia32_vfnmaddpd256_mask,
3829         __builtin_ia32_vfnmaddpd256_maskz,
3830         __builtin_ia32_vfnmaddpd128_mask,
3831         __builtin_ia32_vfnmaddpd128_maskz,
3832         __builtin_ia32_vfnmaddps256_mask,
3833         __builtin_ia32_vfnmaddps256_maskz,
3834         __builtin_ia32_vfnmaddps128_mask,
3835         __builtin_ia32_vfnmaddps128_maskz,
3836         __builtin_ia32_vfnmaddpd512_mask,
3837         __builtin_ia32_vfnmaddpd512_maskz,
3838         __builtin_ia32_vfnmaddps512_mask,
3839         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3840         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3841         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3842         __builtin_ia32_vfnmaddpd256.
3843         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3844         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3845         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3846         Likewise.
3847         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3848         Likewise.
3849         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3850         Likewise.
3851         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3853 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3855         PR target/72782
3856         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3857         __builtin_ia32_vfmsubpd512_mask.
3858         (_mm512_mask_fmsub_round_pd): Likewise.
3859         (_mm512_fmsub_pd): Likewise.
3860         (_mm512_mask_fmsub_pd): Likewise.
3861         (_mm512_maskz_fmsub_round_pd): Use
3862         __builtin_ia32_vfmsubpd512_maskz.
3863         (_mm512_maskz_fmsub_pd): Likewise.
3864         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3865         (_mm512_mask_fmsub_round_ps): Likewise.
3866         (_mm512_fmsub_ps): Likewise.
3867         (_mm512_mask_fmsub_ps): Likewise.
3868         (_mm512_maskz_fmsub_round_ps): Use
3869         __builtin_ia32_vfmsubps512_maskz.
3870         (_mm512_maskz_fmsub_ps): Likewise.
3871         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3872         __builtin_ia32_vfmsubpd256_mask.
3873         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3874         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3875         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3876         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3877         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3878         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3879         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3880         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3881         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3882         __builtin_ia32_vfmsubpd.
3883         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3884         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3885         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3886         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3887         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3888         * config/i386/i386-builtin.def: Add
3889         __builtin_ia32_vfmsubpd256_mask,
3890         __builtin_ia32_vfmsubpd256_maskz,
3891         __builtin_ia32_vfmsubpd128_mask,
3892         __builtin_ia32_vfmsubpd128_maskz,
3893         __builtin_ia32_vfmsubps256_mask,
3894         __builtin_ia32_vfmsubps256_maskz,
3895         __builtin_ia32_vfmsubps128_mask,
3896         __builtin_ia32_vfmsubps128_maskz,
3897         __builtin_ia32_vfmsubpd512_mask,
3898         __builtin_ia32_vfmsubpd512_maskz,
3899         __builtin_ia32_vfmsubps512_mask,
3900         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3901         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3902         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3903         __builtin_ia32_vfmsubpd256.
3904         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3905         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3906         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3907         Likewise.
3908         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3909         Likewise.
3910         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3911         Likewise.
3912         (fmai_vmfmsub_<mode><round_name>): Likewise.
3914 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3916         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3917         Remove plus.  Renamed to ...
3918         (*sub<mode>3<mask_name>_bcst): This.
3919         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3920         (*add<mode>3<mask_name>_bcst): This.
3922 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3924         PR target/72782
3925         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3927 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3929         PR target/87662
3930         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3931         (_mm_or_epi32): Likewise.
3932         (_mm256_xor_epi32): Likewise.
3933         (_mm_xor_epi32): Likewise.
3934         (_mm256_or_epi64): Likewise.
3935         (_mm_or_epi64): Likewise.
3936         (_mm256_xor_epi64): Likewise.
3937         (_mm_xor_epi64): Likewise.
3939 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
3941         PR target/72782
3942         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3944 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
3946         PR middle-end/87647
3947         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3949 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
3951         * doc/ux.texi: Move @section directly after @node.
3953 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
3955         PR middle-end/85488
3956         PR middle-end/87649
3957         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3958         depend closely nested inside of loop with ordered clause with
3959         a parameter.
3961 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
3963         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3964         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3965         * doc/ux.texi: New file.
3967 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
3969         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3970         be the first CR field allocated.
3972 2018-10-19  Richard Biener  <rguenther@suse.de>
3974         PR target/87657
3975         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3976         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3978 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3980         PR target/72782
3981         * config/i386/sse.md
3982         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3983         (*add<mode>3<mask_name>_bcst_2): Likewise.
3985 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
3987         * config/i386/sse.md
3988         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3989         Replace nonimmediate_operand with register_operand.
3990         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3991         Likewise.
3992         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3993         Likewise.
3995 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3997         PR rtl-optimization/87596
3998         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3999         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4000         for instructions in FROM..TO range.
4002 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4004         * cfgexpand.c (expand_one_var): Use specific wording in error message
4005         for non-local frame variables.
4006         * stor-layout.c (layout_decl): Do not issue a warning for them.
4008 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4010         * haifa-sched.c (priority): Add force_recompute parameter.
4011         (apply_replacement): Call priority () with force_recompute = true.
4012         (restore_pattern): Likewise.
4014 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4016         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4017         HOST_BITS_PER_WIDE_INT.
4018         (test_vector_ops_duplicate): Likewise.
4020 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4022         PR target/72782
4023         * config/i386/sse.md (VF_AVX512): New.
4024         (avx512bcst): Likewise.
4025         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4026         Likewise.
4027         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4028         Likewise.
4029         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4030         Likewise.
4032 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4034         * doc/invoke.texi (-dumpversion): Improve grammar.
4035         (-dumpfullversion): Make more consistent with -dumpversion.
4037 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4039         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4040         Set JUMP_LABEL to the jump insn.
4041         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4042         Predict emitted jump and add label to jump insn.
4044 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4046         PR tree-optimization/87562
4047         * input.c (get_substring_ranges_for_loc): Use
4048         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4049         getting the linemap for the endpoint.  Verify that it's either
4050         in the same linemap as the start point's spelling location, or
4051         at least in the same file.
4053 2018-10-18  Richard Biener  <rguenther@suse.de>
4055         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4056         feed width-specific load/store costs through ix86_vec_cost.
4057         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4058         (k8_cost): Likewise.
4059         (bdver_cost): Likewise.
4060         (znver1_cost): Likewise.
4061         (btver1_cost): Likewise.
4062         (btver2_cost): Likewise.
4064 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4066         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4067         to simplify subreg of vec_merge.
4069 2018-10-18  Richard Biener  <rguenther@suse.de>
4071         * config/i386/i386.c: Fix costing of vector FMA.
4073 2018-10-18  Richard Biener  <rguenther@suse.de>
4075         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4076         and argument.
4077         (ix86_builtin_vectorization_cost): For vec_construct properly
4078         cost insertion into SSE regs.
4079         (...): Adjust calls to ix86_vec_cost.
4081 2018-10-18  Richard Biener  <rguenther@suse.de>
4083         PR middle-end/87087
4084         Revert
4085         2018-02-07  Richard Biener  <rguenther@suse.de>
4087         PR tree-optimization/84204
4088         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4089         this place.
4091 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4093         PR target/87537
4094         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4095         of vec_duplicate.
4096         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4097         VEC_MERGE of a VEC_DUPLICATE.
4099 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4101         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4102         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4103         * doc/standards.texi (C Language): Document C2X.
4104         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4105         (rl78_option_override): Handle "GNU C2X" language name.
4107 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4109         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4110         Document C17 as published in 2018.
4112 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4114         PR middle-end/87623
4115         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4116         bail out if both sides do not have the same storage order.
4118 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4120         * bitmap.c (bitmap_head::dump): New.
4121         * bitmap.h (bitmap_head): Add dump().
4122         * gimple-ssa-evrp-analyze.c
4123         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4124         (evrp_range_analyzer::set_ssa_range_info): Same.
4125         (evrp_range_analyzer::record_ranges_from_phis): Same.
4126         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4127         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4128         * gimple-ssa-sprintf.c (get_int_range): Same.
4129         (format_integer): Same.
4130         (sprintf_dom_walker::handle_gimple_call): Same.
4131         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4132         (ipcp_vr_lattice::top_p): Same.
4133         (ipcp_vr_lattice::bottom_p): Same.
4134         (ipcp_vr_lattice::set_to_bottom): Same.
4135         (ipa_vr_operation_and_type_effects): Same.
4136         (propagate_vr_across_jump_function): Same.
4137         (ipcp_store_vr_results): Same.
4138         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4139         (ipa_print_node_jump_functions_for_edge): Same.
4140         (ipa_get_value_range): Same.
4141         (ipa_compute_jump_functions_for_edge): Same.
4142         (ipa_write_jump_function): Same.
4143         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4144         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4145         Same.
4146         * vr-values.c (set_value_range_to_nonnegative): Same.
4147         (set_value_range_to_truthvalue): Same.
4148         (vr_values::get_value_range): Same.
4149         (vr_values::set_defs_to_varying): Same.
4150         (vr_values::update_value_range): Same.
4151         (symbolic_range_based_on_p): Same.
4152         (vr_values::op_with_boolean_value_range_p): Same.
4153         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4154         (vr_values::extract_range_from_ssa_name): Same.
4155         (vr_values::extract_range_from_binary_expr): Same.
4156         (vr_values::extract_range_from_unary_expr): Same.
4157         (vr_values::extract_range_from_cond_expr): Same.
4158         (vr_values::extract_range_from_comparison): Same.
4159         (vr_values::check_for_binary_op_overflow): Same.
4160         (vr_values::extract_range_basic): Same.
4161         (vr_values::extract_range_from_assignment): Same.
4162         (compare_ranges): Same.
4163         (compare_range_with_value): Same.
4164         (vr_values::adjust_range_with_scev): Same.
4165         (vrp_valueize): Same.
4166         (vrp_valueize_1): Same.
4167         (vr_values::get_vr_for_comparison): Same.
4168         (vr_values::compare_name_with_value): Same.
4169         (vr_values::compare_names): Same.
4170         (vr_values::vrp_evaluate_conditional): Same.
4171         (find_case_label_ranges): Same.
4172         (vr_values::vrp_visit_switch_stmt): Same.
4173         (vr_values::extract_range_from_phi_node): Same.
4174         (vr_values::simplify_div_or_mod_using_ranges): Same.
4175         (vr_values::simplify_bit_ops_using_ranges): Same.
4176         (test_for_singularity): Same.
4177         (range_fits_type_p): Same.
4178         (vr_values::simplify_cond_using_ranges_1): Same.
4179         (vr_values::simplify_switch_using_ranges): Same.
4180         (vr_values::simplify_float_conversion_using_ranges): Same.
4181         (vr_values::two_valued_val_range_p): Same.
4182         (vr_values::add_equivalence): Move to value_range::equiv_add.
4183         * vr-values.h (vr_values::add_equivalence): Remove.
4184         (VR_INITIALIZER): Remove.
4185         * tree-vrp.c (value_range::set): New.
4186         (value_range::equiv_add): New.
4187         (value_range::value_range): New.
4188         (value_range::deep_copy): New.
4189         (value_range::check): New.
4190         (value_range::equal_p): New.
4191         (value_range::ignore_equivs_equal_p): New.
4192         (value_range::operator==): New.
4193         (value_range::operator!=): New.
4194         (value_range::symbolic_p): New.
4195         (value_range::numeric_p): New.
4196         (value_range::set_undefined): New.
4197         (value_range::set_varying): New.
4198         (value_range::may_contain_p): New.
4199         (value_range::equiv_clear): New.
4200         (value_range::singleton_p): New.
4201         (value_range::intersect): New.
4202         (value_range::dump): New.
4203         (value_range::set_and_canonicalize): New.
4204         (set_value_range): Adjust for value_range API.
4205         (set_value_range_to_undefined): Same.
4206         (set_value_range_to_varying): Same.
4207         (set_and_canonicalize_value_range): Same.
4208         (set_value_range_to_nonnull): Same.
4209         (set_value_range_to_null): Same.
4210         (range_is_null): Same.
4211         (range_is_nonnull): Same.
4212         (range_int_cst_p): Same.
4213         (range_int_cst_singleton_p): Same.
4214         (symbolic_range_p): Same.
4215         (range_includes_zero_p): Same.
4216         (value_range_constant_singleton): Same.
4217         (vrp_set_zero_nonzero_bits): Same.
4218         (ranges_from_anti_range): Same.
4219         (extract_range_into_wide_ints): Same.
4220         (extract_range_from_multiplicative_op): Same.
4221         (set_value_range_with_overflow): Same.
4222         (extract_range_from_binary_expr_1): Same.
4223         (extract_range_from_unary_expr): Same.
4224         (dump_value_range): Same.
4225         (debug_value_range): Same.
4226         (vrp_prop::check_array_ref): Same.
4227         (vrp_prop::check_mem_ref): Same.
4228         (vrp_prop::vrp_initialize): Same.
4229         (vrp_prop::visit_stmt): Same.
4230         (intersect_ranges): Same.
4231         (vrp_prop::visit_phi): Same.
4232         (vrp_prop::vrp_finalize): Same.
4233         (determine_value_range_1): Same.
4234         (determine_value_range): Same.
4235         (vrp_intersect_ranges_1): Rename to...
4236         (vrp_intersect_1): this.
4237         (vrp_intersect_ranges): Rename to...
4238         (value_range::intersect_helper): ...this.
4239         (vrp_meet_1): Rename to...
4240         (value_range::union_helper): ...this.
4241         (vrp_meet): Rename to...
4242         (value_range::union_): ...this.
4243         (copy_value_range): Remove.
4244         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4245         (value_range::vrtype): New.
4246         (value_range::type): New.
4247         (value_range::equiv): New.
4248         (value_range::min): New.
4249         (value_range::max): New.
4250         (value_range::varying_p): New.
4251         (value_range::undefined_p): New.
4252         (value_range::null_p): New.
4253         (value_range::equiv_add): New.
4254         (copy_value_range): Remove.
4256 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4258         * Makefile.in (SELFTEST_TARGETS): New.
4259         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4260         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4261         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4262         c/Make-lang.in.
4263         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4264         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4265         cp/Make-lang.in.
4266         * configure: Regenerate.
4267         * configure.ac (selftest_languages): New.
4269 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4271         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4272         overflow wraps argument.
4273         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4274         wide_int_range_multiplicative_op.
4275         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4276         overflow wraps argument.
4277         (wide_int_range_multiplicative_op): Same.
4278         (wide_int_range_lshift): Same.
4279         (wide_int_range_div): Same.
4280         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4281         (wide_int_range_lshift): Same.
4282         (wide_int_range_div): Same.
4284 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4286         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4287         use sign as argument.
4288         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4289         wide_int_range_shift_undefined_p.
4291 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4293         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4294         Rename to...
4295         (@despeculate_copy<ALLI_TI:mode>): ... This.
4296         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4297         switch statement.
4299 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4301         * config.gcc: Obsolete *-*-solaris2.10*.
4302         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4304 2018-10-12  Jeff Law  <law@redhat.com>
4306         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4307         reg + sym +- const_int addressing modes.
4309 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4311         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4312         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4313         margin width.
4314         (layout::start_annotation_line): Only print up to 3 of the
4315         margin character, to avoid touching the left-hand side.
4316         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4317         minimum margin width, as set by test_diagnostic_context's ctor.
4318         (selftest::test_fixit_insert_containing_newline): Likewise.
4319         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4320         (selftest::test_line_numbers_multiline_range): Clear
4321         dc.min_margin_width.
4322         * diagnostic.c (diagnostic_initialize): Initialize
4323         min_margin_width.
4324         * diagnostic.h (struct diagnostic_context): Add field
4325         "min_margin_width".
4326         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4327         * opts.c (common_handle_option): Handle
4328         OPT_fdiagnostics_minimum_margin_width_.
4329         * selftest-diagnostic.c
4330         (selftest::test_diagnostic_context::test_diagnostic_context):
4331         Initialize min_margin_width to 6.
4332         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4334 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4336         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4337         Fix usage of "error_at_rich_loc" in the comment.
4339 2018-10-15  Renlin Li  <renlin.li@arm.com>
4341         PR target/87563
4342         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4343         if-conversioned loop when it contains ifn with types not
4344         supported by backend.
4345         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4346         (direct_internal_fn_supported_p): New helper function.
4347         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4349 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4351         PR target/87572
4352         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4353         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4354         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4356 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4358         PR tree-optimization/87022
4359         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4360         bits in dist vector rather than the first one.
4362 2018-10-15  Richard Biener  <rguenther@suse.de>
4364         PR middle-end/87610
4365         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4366         (visit_loadstore): When a used restrict tag escaped verify that
4367         the points-to solution of "other" pointers do not include
4368         escaped.
4369         (compute_dependence_clique): If a used restrict tag escaped
4370         communicated that down to visit_loadstore.
4372 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4374         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4375         into reg if it isn't a general operand.
4377 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4379         PR target/87599
4380         * config/i386/sse.md (*vec_dupv2di): Add register source to
4381         movddup.
4383 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4385         PR target/87572
4386         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4387         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4388         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4389         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4391 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4393         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4394         (notice_args_size): Set it in the current trace if no insn that can
4395         throw internally has been seen yet.
4396         (connect_traces): When connecting args_size between traces, allow the
4397         incoming values not to match if there is an insn setting it before the
4398         first insn that can throw internally; in that case, force the creation
4399         of a CFI note on this latter insn.
4401 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4403         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4404         the base class, because using the injected-class-name was not clearly
4405         specified until DR 176.
4407 2018-10-12  Paul Koning  <ni1d@arrl.net>
4409         * config/pdp11/pdp11.md (doloop_end): New expander.
4410         (doloop_end_insn): renamed from "doloop_end".
4411         (addqi3): New pattern.
4412         (subqi3): New pattern.
4413         * config/pdp11/predicates.md (incdec_operand): New predicate.
4415 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4417         PR middle-end/81376
4418         * real.c (format_helper::can_represent_integral_type_p): New function
4419         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4420         * match.pd: New pattern.
4422 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4424         * configure.ac: Introduce --enable-large-address-aware
4425         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4426         * doc/install.texi: Document it.
4427         * configure, config.in: Rebuilt.
4428         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4429         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4430         (LINK_SPEC): Insert it.
4431         * config/i386/mingw-w64.h: Likewise.
4433         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4435 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4437         PR rtl-optimization/87600
4438         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4440 2018-10-12  Paul Koning  <ni1d@arrl.net>
4442         * doc/md.texi (doloop_end): Document that the pattern code may
4443         need to check operand mode.
4445 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4447         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4448         to zero-extend between int and floating-point registers.
4449         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4450         ldp into floating-point registers.  Add type and arch attributes.
4451         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4452         Use f_loads for type attribute.
4454 2018-10-11  Martin Sebor  <msebor@redhat.com>
4456         * doc/extend.texi (attribute packed): Correct typos.
4458 2018-10-11  Martin Sebor  <msebor@redhat.com>
4460         * doc/extend.texi (attribute flatten): Mention interaction with
4461         noinline.
4463 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4465         PR target/87156
4466         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4467         Set new_decl virtual flag to zero.
4469 2018-10-11  Martin Sebor  <msebor@redhat.com>
4471         PR middle-end/87593
4472         * doc/extend.texi (attribute format_arg): Discuss using multiple
4473         attributes on a single function.
4475 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4477         PR tree-optimization/86829
4478         * match.pd (sin (atan (x))): New simplification rules.
4479         (cos (atan (x))): Likewise.
4480         * real.c (build_sinatan_real): New function.
4481         * real.h (build_sinatan_real): Prototype.
4483 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4485         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4486         function.
4487         (fold_mergeeo_helper): New helper function.
4488         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4489         intrinsics.  Correct some whitespace indentation issues.
4491 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4493         PR target/87511
4494         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4495         Use HOST_WIDE_INT_1U for shift.
4497 2018-10-11  Doug Rupp  <rupp@adacore.com>
4498             Olivier Hainque  <hainque@adacore.com>
4500         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4501         Pass --relax to the linker for RTPs.
4502         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4504 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4505             Jan Hubicka  <jh@suse.cz>
4506             Martin Jambor  <mjambor@suse.cz>
4508         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4509         the same elements are repeated rather than printing all of them.
4510         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4511         "repeated" elements.
4512         * read-rtl-function.c (test_loading_repeat): New function.
4513         (read_rtl_function_c_tests): Call test_loading_repeat.
4514         * rtl-tests.c (test_dumping_repeat): New function.
4515         (rtl_tests_c_tests): Call test_dumping_repeat.
4517 2018-10-11  Richard Biener  <rguenther@suse.de>
4519         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4520         bdver?_cost): Unify to ...
4521         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4522         * config/i386/i386.c (processor_cost_table): Adjust.
4524 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4526         PR middle-end/87574
4527         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4528         the thunk when expanding to GIMPLE.
4530 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4532         * varasm.c (mergeable_string_section): Don't try to move zero-length
4533         strings to the merge section.
4535 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4537         PR target/87573
4538         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4540 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4542         PR target/87550
4543         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4544         to special_args set.
4546 2018-10-10  Richard Biener  <rguenther@suse.de>
4548         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4549         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4550         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4551         and recursing and pattern terminating the recursion on SSE
4552         vector width using ix86_expand_reduc.
4553         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4554         width and recursing and SSE2 vector variant doing the final
4555         reduction with ix86_expand_reduc.
4556         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4557         with terminating the recursion at AVX level, splitting that
4558         to SSE there.
4560 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4562         * genmatch.c (error_cb): Rename to...
4563         (diagnostic_cb): ...this, converting int params to enums.
4564         (fatal_at): Update for renaming.
4565         (warning_at): Likewise.
4566         (main): Likewise.
4567         * input.c (selftest::ebcdic_execution_charset::apply):
4568         Update for renaming of...
4569         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4570         to...
4571         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4572         converting level and reason to enums.
4573         (class selftest::lexer_error_sink): Rename to...
4574         (class selftest::lexer_test_options): ...this, renaming field
4575         "m_errors" to "m_diagnostics".
4576         (selftest::lexer_test_options::apply): Update for renaming of...
4577         (selftest::lexer_test_options::on_error): ...this, renaming to...
4578         (selftest::lexer_test_options::on_diagnostic): ...this
4579         converting level and reason to enums.
4580         (selftest::test_lexer_string_locations_raw_string_unterminated):
4581         Update for renamings.
4582         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4583         "reason".
4585 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4587         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4588         * config/rs6000/pmmintrin.h: New file.
4590 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4592         PR tree-optimization/86659
4593         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4595 2018-10-09  Richard Biener  <rguenther@suse.de>
4597         PR tree-optimization/63155
4598         * tree-ssa-structalias.c: Include tree-ssa.h.
4599         (get_constraint_for_ssa_var): For undefs return nothing_id.
4600         (find_func_aliases): Cleanup PHI handling.
4602 2018-10-09  Richard Biener  <rguenther@suse.de>
4604         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4605         replacements.
4607 2018-10-09  Martin Liska  <mliska@suse.cz>
4609         * asan.c (asan_emit_stack_protection): If a stack variable
4610         is located in a same file as current function, then emit
4611         line info into variable definition string.
4613 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4615         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4616         information.
4618 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4620         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4621         on the thunk.
4623 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4625         PR tree-optimization/86659
4626         * gimple-match.h (struct gimple_match_op): Add reverse field.
4627         (gimple_match_op::set_op): New overloaded method.
4628         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4629         the REF_REVERSE_STORAGE_ORDER flag on the value.
4630         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4631         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4632         
4633 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4635         PR middle-end/63155
4636         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4637         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4639 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4641         PR target/87517
4642         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4643         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4645 2018-10-08  Richard Biener  <rguenther@suse.de>
4647         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4648         cost the same as AVX128 ones.
4650 2018-10-08  Paul Koning  <ni1d@arrl.net>
4652         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4653         (expand_block_move): New function.
4654         * config/pdp11/pdp11.c (output_block_move): Remove.
4655         (expand_block_move): New function.
4656         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4657         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4658         (*movmemhi1): Remove.
4660 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4662         * config/s390/2827.md: Increase latencies for some FP instructions.
4664 2018-10-08  Richard Biener  <rguenther@suse.de>
4666         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4667         Open a dump scope.
4668         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4669         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4670         (add_stmt_cost): Dump return value of the hook.
4672 2018-10-08  Richard Biener  <rguenther@suse.de>
4674         PR tree-optimization/63155
4675         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4676         (ssa_propagation_engine::ssa_propagate): Remove redundant
4677         bitmap bit clearing.
4679 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4681         PR rtl-optimization/86939
4682         PR rtl-optimization/87479
4683         * ira.h (non_conflicting_reg_copy_p): New prototype.
4684         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4685         (make_hard_regno_dead): Don't add conflicts for register
4686         ignore_reg_for_conflicts.
4687         (make_object_dead): Likewise.
4688         (non_conflicting_reg_copy_p): New function.
4689         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4690         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4691         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4692         (make_hard_regno_dead): Don't add conflicts for register
4693         ignore_reg_for_conflicts.  Remove special conflict handling of
4694         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4695         check_pic_pseudo_p and update callers.
4696         (mark_pseudo_dead): Don't add conflicts for register
4697         ignore_reg_for_conflicts.
4698         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4700 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4701             Jim Wilson  <jimw@sifive.com>
4703         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4704         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4705         new pattern using HONOR_SNANS that emits one extra instruction.
4707 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4709         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4710         patterns): Merge SI and DI patterns to a GPR pattern.
4711         (unnamed define_insn and define_split for record form of that): Merge
4712         to a single define_insn_and_split pattern.
4714 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4716         PR c++/56856
4717         * input.c (expand_location_to_spelling_point): Add param "aspect"
4718         and use rather than hardcoding LOCATION_ASPECT_CARET.
4719         (get_substring_ranges_for_loc): Handle the case of a single token
4720         within a macro expansion.
4721         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4722         defaulting to LOCATION_ASPECT_CARET.
4724 2018-10-05  Paul Koning  <ni1d@arrl.net>
4726         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4727         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4728         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4729         (pdp11_guard_type): New function.
4731 2018-10-05  Paul Koning  <ni1d@arrl.net>
4733         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4734         * config/pdp11/pdp11.opt (mfloat32): Remove.
4735         (mfloat64): Remove.
4736         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4737         (pdp11 -mfloat64): Remove.
4739 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4741         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4742         (*cmp<mode>_cc_i387): Ditto.
4743         (*cmpu<mode>_cc_i387): Ditto.
4744         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4745         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4746         "scratch" argument.
4747         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4748         Emit x86_sahf_1 pattern.
4749         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4750         (ix86_expand_carry_flag_compare): Ditto.
4752 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4754         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4755         to reg_or_0_operand.  Add "C" constraint.
4756         (*cmpxf_cc_i387): Ditto.
4757         (*cmp<mode>_i387): Change operand 2 predicate
4758         to nonimm_or_0_operand.  Add "C" constraint.
4759         (*cmp<mode>_cc_i387): Ditto.
4760         (*cmp<mode>_0_i387): Remove insn pattern.
4761         (*cmp<mode>_0_cc_i387): Ditto.
4763 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4765         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4766         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4767         from vector_move_operand.  Update all uses.
4769 2018-10-05  Martin Sebor  <msebor@redhat.com>
4771         PR tree-optimization/87490
4772         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4773         consistently.
4775 2018-10-05  Richard Biener  <rguenther@suse.de>
4777         PR tree-optimization/63155
4778         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4779         vertical space in dumpfiles.
4780         * tree-ssa-propagate.h
4781         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4782         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4783         (ssa_edge_worklist_back): Likewise.
4784         (curr_order): Likewise.
4785         (cfg_blocks_get): Remove abstraction.
4786         (cfg_blocks_add): Likewise.
4787         (cfg_blocks_empty_p): Likewise.
4788         (add_ssa_edge): Add to current or next worklist based on
4789         RPO index.
4790         (add_control_edge): Likewise.
4791         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4792         into ...
4793         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4794         iteration from CFG and SSA edge worklist so we process
4795         everything in RPO order, prioritizing forward progress
4796         over iteration.
4797         (ssa_prop_init): Allocate new worklists, do not dump
4798         immediate uses.
4799         (ssa_prop_fini): Free new worklists.
4801 2018-10-05  Richard Biener  <rguenther@suse.de>
4803         * tree-core.h (tree_block::abstract_flag): Remove.
4804         (tree_block::block_num): Make full 32bits.
4805         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4806         * tree.h (BLOCK_ABSTRACT): Remove.
4807         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4808         resulting from BLOCK_ABSTRACT being always false.
4809         (gen_inlined_subroutine_die): Likewise.
4810         (gen_block_die): Likewise.
4811         * tree.c (block_ultimate_origin): Likewise.
4812         * tree-pretty-print.c (dump_block_node): Remove code dealing
4813         with BLOCK_ABSTRACT.
4814         * tree-ssa-live.c (dump_scope_block): Likewise.
4815         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4816         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4818 2018-10-05   Richard Biener  <rguenther@suse.de>
4820         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4821         is asked for initialize mode to the component mode of the
4822         vector type.
4824 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4826         PR target/87522
4827         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4828         assembler for -mavx.
4829         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4831 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4833         PR target/87509
4834         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4835         RS6000_BTM_DFP.
4836         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4837         to be DImode.  When using mffscrn, force the operand to a register.
4839 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4841         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4842         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4843         X87MODEF mode iterator.
4844         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4845         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4846         X87MODEF mode iterator.
4848 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4850         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4851         -Wno-prio-ctor-dtor.
4853 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4855         * Makefile.in (OBJS): Add opt-problem.o.
4856         * dump-context.h: Include "selftest.h.
4857         (selftest::temp_dump_context): New forward decl.
4858         (class dump_context): Make friend of class
4859         selftest::temp_dump_context.
4860         (dump_context::dump_loc_immediate): New decl.
4861         (class dump_pretty_printer): Move here from dumpfile.c.
4862         (class temp_dump_context): Move to namespace selftest.
4863         (temp_dump_context::temp_dump_context): Add param
4864         "forcibly_enable_dumping".
4865         (selftest::verify_dumped_text):
4866         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4867         (selftest::verify_item):
4868         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4869         (ASSERT_IS_TREE): Likewise.
4870         (ASSERT_IS_GIMPLE): Likewise.
4871         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4872         to...
4873         (dump_context::dump_loc_immediate): ...this new function.
4874         (class dump_pretty_printer): Move to dump-context.h.
4875         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4876         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4877         (temp_dump_context::temp_dump_context): Move to "selftest"
4878         namespace.  Add param "forcibly_enable_dumping", and use it to
4879         conditionalize the use of m_pp;
4880         (selftest::verify_dumped_text): Make non-static.
4881         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4882         (selftest::verify_item): Make non-static.
4883         (ASSERT_IS_TEXT): Move to dump-context.h.
4884         (ASSERT_IS_TREE): Likewise.
4885         (ASSERT_IS_GIMPLE): Likewise.
4886         (selftest::test_capture_of_dump_calls): Pass "true" for new
4887         param of temp_dump_context.
4888         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4889         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
4890         TDF_COMPARE_DEBUG.
4891         * opt-problem.cc: New file.
4892         * opt-problem.h: New file.
4893         * optinfo-emit-json.cc
4894         (selftest::test_building_json_from_dump_calls): Pass "true" for
4895         new param of temp_dump_context.
4896         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4897         (optinfo::emit_for_opt_problem): New function.
4898         (optinfo::emit): Clarity which emit_item is used.
4899         * optinfo.h (optinfo::get_dump_location): New accessor.
4900         (optinfo::emit_for_opt_problem): New decl.
4901         (optinfo::emit): Make const.
4902         * selftest-run-tests.c (selftest::run_tests): Call
4903         selftest::opt_problem_cc_tests.
4904         * selftest.h (selftest::opt_problem_cc_tests): New decl.
4905         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4906         bool to opt_result, converting fprintf messages to
4907         opt_result::failure_at calls.  Add "stmt" param for use by the
4908         failure_at calls.
4909         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4910         (runtime_alias_check_p): Convert return type from bool to
4911         opt_result, converting dump_printf calls to
4912         opt_result::failure_at, using the statement DDR_A for their
4913         location.
4914         (find_data_references_in_stmt): Convert return type from bool to
4915         opt_result, converting "return false" to opt_result::failure_at
4916         with a new message.
4917         * tree-data-ref.h: Include "opt-problem.h".
4918         (dr_analyze_innermost): Convert return type from bool to opt_result,
4919         and add a const gimple * param.
4920         (find_data_references_in_stmt): Convert return type from bool to
4921         opt_result.
4922         (runtime_alias_check_p): Likewise.
4923         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4924         dr_analyze_innermost.
4925         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4926         Convert return type from bool to opt_result, adding a message for
4927         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4928         (vect_analyze_data_ref_dependence): Convert return type from bool
4929         to opt_result.  Change sense of return type from "false"
4930         effectively meaning "no problems" to "false" meaning a problem,
4931         so that "return false" becomes "return opt_result::success".
4932         Convert "return true" calls to opt_result::failure_at, using
4933         the location of statement A rather than vect_location.
4934         (vect_analyze_data_ref_dependences): Convert return type from bool
4935         to opt_result.
4936         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4937         calls to opt_result::failure_at, using the stmt location rather
4938         than vect_location.
4939         (vect_verify_datarefs_alignment): Convert return type from bool
4940         to opt_result.
4941         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
4942         into multiple more-tightly-scoped copies.
4943         (vect_analyze_data_refs_alignment): Convert return type from bool
4944         to opt_result.
4945         (vect_analyze_data_ref_accesses): Likewise, converting a
4946         "return false" to a "return opt_result::failure_at", adding a
4947         new message.
4948         (vect_prune_runtime_alias_test_list): Convert return type from
4949         bool to opt_result, converting dump_printf_loc to
4950         opt_result::failure_at.  Add a %G to show the pertinent statement,
4951         and use the stmt's location rather than vect_location.
4952         (vect_find_stmt_data_reference): Convert return type from
4953         bool to opt_result, converting dump_printf_loc to
4954         opt_result::failure_at, using stmt's location.
4955         (vect_analyze_data_refs):  Convert return type from bool to
4956         opt_result.  Convert "return false" to "return
4957         opt_result::failure_at", adding messages as needed.
4958         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4959         type from bool to opt_result.
4960         (vect_determine_vf_for_stmt): Likewise.
4961         (vect_determine_vectorization_factor): Likewise, converting
4962         dump_printf_loc to opt_result::failure_at, using location of phi
4963         rather than vect_location.
4964         (vect_analyze_loop_form_1): Convert return type from bool to
4965         opt_result, converting dump_printf_loc calls, retaining the use of
4966         vect_location.
4967         (vect_analyze_loop_form): Convert return type from loop_vec_info
4968         to opt_loop_vec_info.
4969         (vect_analyze_loop_operations): Convert return type from bool to
4970         opt_result, converting dump_printf_loc calls, using the location
4971         of phi/stmt rather than vect_location where available.  Convert
4972         various "return false" to "return opt_result::failure_at" with
4973         "unsupported phi" messages.
4974         (vect_get_datarefs_in_loop): Convert return type from bool to
4975         opt_result.  Add a message for the
4976         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4977         (vect_analyze_loop_2): Convert return type from bool to
4978         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
4979         each "goto again;", adding new messages where needed.
4980         Add "unsupported grouped {store|load}" messages.
4981         (vect_analyze_loop): Convert return type from loop_vec_info to
4982         opt_loop_vec_info.
4983         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4984         bool to opt_result.
4985         * tree-vect-stmts.c (process_use): Likewise, converting
4986         dump_printf_loc call and using stmt location, rather than
4987         vect_location.
4988         (vect_mark_stmts_to_be_vectorized): Likeise.
4989         (vect_analyze_stmt): Likewise, adding a %G.
4990         (vect_get_vector_types_for_stmt): Convert return type from bool to
4991         opt_result, converting dump_printf_loc calls and using stmt
4992         location, rather than vect_location.
4993         (vect_get_mask_type_for_stmt): Convert return type from tree to
4994         opt_tree, converting dump_printf_loc calls and using stmt location.
4995         * tree-vectorizer.c: Include "opt-problem.h.
4996         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4997         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
4998         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
4999         enabled, use it to report at the top level "couldn't vectorize
5000         loop" followed by the problem.
5001         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5002         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5003         to opt_result.
5004         (vect_analyze_stmt): Likewise.
5005         (vect_get_vector_types_for_stmt): Likewise.
5006         (tree vect_get_mask_type_for_stmt): Likewise.
5007         (vect_analyze_data_ref_dependences): Likewise.
5008         (vect_enhance_data_refs_alignment): Likewise.
5009         (vect_analyze_data_refs_alignment): Likewise.
5010         (vect_verify_datarefs_alignment): Likewise.
5011         (vect_analyze_data_ref_accesses): Likewise.
5012         (vect_prune_runtime_alias_test_list): Likewise.
5013         (vect_find_stmt_data_reference): Likewise.
5014         (vect_analyze_data_refs): Likewise.
5015         (vect_analyze_loop): Convert return type from loop_vec_info to
5016         opt_loop_vec_info.
5017         (vect_analyze_loop_form): Likewise.
5018         (vect_analyze_slp): Convert return type from bool to opt_result.
5020 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5022         * doc/invoke.texi (-fopt-info): Document new "internals"
5023         sub-option.
5024         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5025         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5026         MSG_ALL_KINDS.
5027         (optinfo_verbosity_options): Add "internals".
5028         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5029         (dump_context::apply_dump_filter_p): New member function.
5030         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5031         explicitly masking the dump_kind.
5032         (dump_context::begin_scope): Increment the scope depth first.  Use
5033         apply_dump_filter_p rather than explicitly masking the dump_kind.
5034         (dump_context::emit_item): Use apply_dump_filter_p rather than
5035         explicitly masking the dump_kind.
5036         (dump_dec): Likewise.
5037         (dump_hex): Likewise.
5038         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5039         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5040         (opt_info_switch_p): Update handling of default
5041         MSG_OPTIMIZED_LOCATIONS to cope with default of
5042         MSG_PRIORITY_USER_FACING.
5043         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5044         masking the dump_kind.
5045         (selftest::test_capture_of_dump_calls): Update test_dump_context
5046         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5047         than MSG_ALL.  Generalize scope test to be run at all four
5048         combinations of with/without MSG_PRIORITY_USER_FACING and
5049         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5050         for each of the two values.
5051         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5052         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5053         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5054         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5055         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5056         with MSG_PRIORITY_*.
5057         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5058         dump messages as MSG_PRIORITY_USER_FACING.
5059         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5060         about the interaction with MSG_PRIORITY_*.
5062 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5064         * varasm.c (output_constant): Add new parameter merge_strings.
5065         Make strings properly zero terminated in merge string sections.
5066         (mergeable_string_section): Don't fail if the last char is non-zero.
5067         (assemble_variable_contents): Handle merge string sections.
5068         (assemble_variable): Likewise.
5069         (assemble_constant_contents): Likewise.
5070         (output_constant_def_contents): Likewise.
5071         (output_constructor_array_range,
5072         output_constructor_regular_field): Adjust call to output_constant.
5073         (output_object_block): Adjust call to assemble_constant_contents
5074         and assemble_variable_contents.
5076 2018-10-04  Martin Liska  <mliska@suse.cz>
5078         PR c/87483
5079         * cgraphunit.c (process_function_and_variable_attributes):
5080         Warn about a function with alias attribute and a body.
5082 2018-10-04  Martin Liska  <mliska@suse.cz>
5084         PR ipa/82625
5085         * multiple_target.c (redirect_to_specific_clone): New function.
5086         (ipa_target_clone): Use it.
5087         * tree-inline.c: Fix comment.
5089 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5091         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5092         fields.
5093         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5094         (gcc::dump_manager::register_pass): New member function, adapted
5095         from loop body in gcc::pass_manager::register_pass, adding a
5096         call to update_dfi_for_opt_info.
5097         (gcc::dump_manager::opt_info_enable_passes): Store the
5098         -fopt-info options into the new fields.  Move the loop
5099         bodies into...
5100         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5101         function.
5102         * dumpfile.h (struct opt_pass): New forward decl.
5103         (gcc::dump_manager::register_pass): New decl.
5104         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5105         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5106         "m_optinfo_flags", and "m_optinfo_filename".
5107         * passes.c (gcc::pass_manager::register_pass): Move all of the
5108         dump-handling code to gcc::dump_manager::register_pass.
5110 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5112         PR rtl-optimization/87466
5113         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5114         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5115         * doc/tm.texi: Regenerate.
5116         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5117         * lra-lives.c (process_bb_lives): Likewise.
5118         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5119         Define.
5121 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5123         * params.c (add_params): Fix initialization.
5125 2018-10-04  Martin Liska  <mliska@suse.cz>
5127         PR gcov-profile/84107
5128         * tree-profile.c (init_ic_make_global_vars):
5129         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5130         Come up with new ic_tuple* variables.  Emit
5131         __gcov_indirect_call{,_topn} variables.
5132         (gimple_gen_ic_profiler): Access the variable
5133         and emit gimple.
5134         (gimple_gen_ic_func_profiler): Access
5135         __gcov_indirect_call.callee field.
5136         (gimple_init_gcov_profiler): Use ptr_type_node.
5137         * value-prof.c (gimple_ic): Use ptr_type_node.
5139 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5141         PR tree-optimization/85787
5142         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5143         malloc_candidate_p into this function and add support for
5144         detecting multiple phis.
5145         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5147 2018-10-04  Martin Liska  <mliska@suse.cz>
5149         PR ipa/87491
5150         * ipa-inline.c (inline_to_all_callers_1):
5151         Call ultimate_alias_target for node being inlined.
5153 2018-10-03  Jeff Law  <law@redhat.com>
5155         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5156         target's wchar_t.
5157         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5158         * tree.h (get_typenode_from_name): Prototype.
5160 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5162         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5163         Change operand 2 predicate to nonimmediate_operand.
5164         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5166 2018-10-03  Martin Sebor  <msebor@redhat.com>
5167             Jeff Law  <law@redhat.com>
5169         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5170         initialize it.
5171         (get_string_length): Detect unterminated arrays.
5172         (format_string): Same.
5173         (format_directive): Warn about unterminated arrays.
5174         (handle_gimple_call): Mark statements with no_warning as needed.
5176 2018-10-03  Jim Wilson  <jimw@sifive.com>
5178         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5179         also define __riscv_abi_rve.  Delete trailing white space.
5181 2018-10-03  Paul Koning  <ni1d@arrl.net>
5183         Enable LRA register allocator for PDP11.
5184         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5185         (R): Likewise.
5186         (D): Likewise.
5187         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5188         * config/pdp11/pdp11.opt (-mlra): New option.
5189         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5191 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5193         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5194         (*<absneg:code>extend<mode>xf2): Ditto.
5196 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5198         PR tree-optimization/87415
5199         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5200         precision fields.
5202 2018-10-02  Jeff Law  <law@redhat.com>
5204         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5205         an unterminated string is discovered.  Bubble up range
5206         even for unterminated strings.
5207         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5208         indicates the string was not terminated via NONSTR.
5210 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5212         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5213         pointer conversions.
5214         Do not do anything special for anti-ranges.
5216 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5218         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5219         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5221 2018-10-03  Martin Liska  <mliska@suse.cz>
5223         PR gcov-profile/86109
5224         * coverage.c (coverage_begin_function): Do not
5225         mark lambdas as artificial.
5226         * tree-core.h (struct GTY): Remove tm_clone_flag
5227         and introduce new lambda_function.
5228         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5230 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5232         PR target/87474
5233         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5234         P8_VECTOR and VSX are enabled.
5236 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5238         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5239         0x3907 as CPU model number.
5241 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5243         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5244         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5245         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5246         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5247         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5248         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5249         value from arch12 to z14.
5251 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5253         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5254         (isinfxf2): Ditto.
5255         (isinf<mode>2): Ditto.
5257 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5259         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5260         before emitting fxam.  Perform calculations in XFmode.
5262 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5264         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5266 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5268         PR middle-end/87319
5269         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5270         * tree.c (signed_or_unsigned_type_for): Handle complex.
5272 2018-10-02  Jeff Law  <law@redhat.com>
5274         * gimple-fold.c (get_range_strlen): Remove dead code.
5276 2018-10-02  Martin Sebor  <msebor@redhat.com>
5277             Jeff Law  <law@redhat.com>
5279         * builtins.c (unterminated_array): Add new arguments.
5280         If argument is not terminated, bubble up size and exact
5281         state to callers.
5282         (expand_builtin_strnlen): Detect, avoid expanding
5283         and diagnose unterminated arrays.
5284         (c_strlen): Fill in offset of start of unterminated strings.
5285         * builtins.h (unterminated_array): Update prototype.
5287 2018-10-02  Richard Biener  <rguenther@suse.de>
5289         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5290         of haddv4df, first reduce to SSE width and exploit the fact
5291         that we only need element zero with the reduction result.
5292         (reduc_plus_scal_v2df): Likewise.
5294 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5296         * dojump.h (do_jump): Delete.
5297         (do_jump_1): Likewise.
5298         (split_comparison): Move around.
5299         * dojump.c (do_jump): Make static.
5300         (do_jump_1): Likewise.
5301         (jumpifnot): Move around.
5302         (jumpifnot_1): Likewise.
5303         (jumpif): Likewise.
5304         (jumpif_1): Likewise.
5305         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5307 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5309         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5310         insns in the delay slot and add_insn_after for the jump insn.
5312 2018-10-02  Richard Biener  <rguenther@suse.de>
5314         * tree-inline.c (expand_call_inline): Use the location of
5315         the callee declaration for the inline-entry marker.
5316         * final.c (notice_source_line): Remove special-casing of
5317         NOTE_INSN_INLINE_ENTRY.
5319 2018-10-01  Carl Love  <cel@us.ibm.com>
5321         PR 69431
5322         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5323         (__builtin_mtfsb0): New.
5324         (__builtin_mtfsb1): New.
5325         ( __builtin_set_fpscr_rn): New.
5326         (__builtin_set_fpscr_drn): New.
5327         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5328         (rs6000_expand_set_fpscr_rn_builtin): Add.
5329         (rs6000_expand_set_fpscr_drn_builtin): Add.
5330         (rs6000_expand_builtin): Add case statement entries for
5331         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5332         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5333         RS6000_BUILTIN_MFFSL.
5334         (rs6000_init_builtins): Add ftype initialization and def_builtin
5335         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5336         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5337         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5338         rs6000_mffscdrn): Add define_insn.
5339         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5340         * doc/extend.texi: Add documentation for the builtins.
5342 2018-10-01  Richard Biener  <rguenther@suse.de>
5344         PR tree-optimization/87465
5345         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5346         causing branch miscounts.
5348 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5350         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5351         aarch64_option_default_param):  New.
5352         (params.h): Include.
5353         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5354         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5355         stack-clash protection validation code.
5357 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5359         * params.c (validate_param): New.
5360         (add_params): Use it.
5361         (set_param_value): Refactor param validation into validate_param.
5362         (diagnostic.h): Include.
5363         * diagnostic.h (diagnostic_ready_p): New.
5365 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5367         * params.c (set_param_value):
5368         Add index of parameter being validated.
5369         * common/common-target.def (option_validate_param): New.
5370         * common/common-targhooks.h (default_option_validate_param): New.
5371         * common/common-targhooks.c (default_option_validate_param): New.
5372         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5373         * doc/tm.texi: Regenerate.
5375 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5377         PR target/86486
5378         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5379         Add validation for stack-clash parameters and set defaults.
5381 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5383         PR target/86486
5384         * configure.ac: Add stack-clash-protection-guard-size.
5385         * doc/install.texi: Document it.
5386         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5387         * params.def: Update comment for guard-size.
5388         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5389         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5390         * configure: Regenerate.
5392 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5394         PR target/86486
5395         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5396         STACK_DYNAMIC_OFFSET): New.
5397         * config/aarch64/aarch64.c (aarch64_layout_frame):
5398         Update outgoing args size.
5399         (aarch64_stack_clash_protection_alloca_probe_range,
5400         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5402 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5404         PR target/86486
5405         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5406         probe ranges.
5407         * target.def (stack_clash_protection_alloca_probe_range): New.
5408         (stack_clash_protection_final_dynamic_probe): Remove.
5409         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5410         (default_stack_clash_protection_final_dynamic_probe): Remove.
5411         * targhooks.c: Likewise.
5412         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5413         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5414         * doc/tm.texi: Regenerate.
5416 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5418         PR target/86486
5419         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5420         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5421         aarch64_clamp_to_uimm12_shift): New.
5422         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5423         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5425 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5427         PR target/86486
5428         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5430 2018-10-01  Jeff Law  <law@redhat.com>
5431             Richard Sandiford <richard.sandiford@linaro.org>
5432             Tamar Christina  <tamar.christina@arm.com>
5434         PR target/86486
5435         * config/aarch64/aarch64.md
5436         (probe_stack_range): Add k (SP) constraint.
5437         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5438         STACK_CLASH_MAX_UNROLL_PAGES): New.
5439         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5440         stack probes for stack clash.
5441         (aarch64_allocate_and_probe_stack_space): New.
5442         (aarch64_expand_prologue): Use it.
5443         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5444         (aarch64_sub_sp): Add emit_move_imm optional param.
5446 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5448         PR tree-optimization/87261
5449         * match.pd: Remove trailing whitespace.
5450         Add (x & y) | ~(x | y) -> ~(x ^ y),
5451         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5453 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5455         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5456         constraints.
5457         (ashlsi3_insn): Update instruction constraints.
5458         (ashrsi3_insn): Likewise.
5459         (rotrsi3): Likewise.
5460         (add_shift): Likewise.
5461         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5462         avoids placing in the limm field small constants which, otherwise,
5463         could end into a small instruction.
5465 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5467         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5468         destination register is not odd-even.
5469         (umaddsidi4_split): Likewise.
5471 2018-10-01  Richard Biener  <rguenther@suse.de>
5473         * tree-inline.c (expand_call_inline): Store origin of fn
5474         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5475         * tree.c (block_ultimate_origin): Simplify and do some
5476         checking.
5478 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5480         * config/i386/mmx.md (EMMS): New int iterator.
5481         (emms): New int attribute.
5482         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5483         EMMS int iterator.  Explicitly declare clobbers.
5484         (mmx_emms): Remove expander.
5485         (mmx_femms): Ditto.
5486         * config/i386/predicates.md (emms_operation): Remove predicate.
5487         (vzeroall_pattern): New predicate.
5488         (vzeroupper_pattern): Rename from vzeroupper_operation.
5489         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5490         vzeroupper_pattern and vzeroall_pattern predicates.
5492 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5494         PR rtl-optimization/86939
5495         * ira-lives.c (make_hard_regno_born): Rename from this...
5496         (make_hard_regno_live): ... to this.  Remove update to conflict
5497         information.  Update function comment.
5498         (make_hard_regno_dead): Add conflict information update.  Update
5499         function comment.
5500         (make_object_born): Rename from this...
5501         (make_object_live): ... to this.  Remove update to conflict information.
5502         Update function comment.
5503         (make_object_dead):  Add conflict information update.  Update function
5504         comment.
5505         (mark_pseudo_regno_live): Call make_object_live.
5506         (mark_pseudo_regno_subword_live): Likewise.
5507         (mark_hard_reg_dead): Update function comment.
5508         (mark_hard_reg_live): Call make_hard_regno_live.
5509         (process_bb_node_lives): Likewise.
5510         * lra-lives.c (make_hard_regno_born): Rename from this...
5511         (make_hard_regno_live): ... to this.  Remove update to conflict
5512         information.  Remove now uneeded check_pic_pseudo_p argument.
5513         Update function comment.
5514         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5515         to conflict information.  Update function comment.
5516         (mark_pseudo_live): Remove update to conflict information.  Update
5517         function comment.
5518         (mark_pseudo_dead): Add conflict information update.
5519         (mark_regno_live): Call make_hard_regno_live.
5520         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5521         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5523 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5525         PR target/87370
5526         * config/i386/i386.c (construct_container): Use TImode for
5527         BLKmode values in 2 integer registers.
5529 2018-09-29  Jeff Law  <law@redhat.com>
5531         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5532         c_strlen rather than just a tree *.
5533         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5534         Update recursive calls appropriately.  If caller did not provide a
5535         suitable data pointer, create a local one.  When a non-terminated
5536         string is discovered, bubble up information about the string via the
5537         c_strlen_data object.
5538         * builtins.h (c_strlen): Update prototype.
5539         (c_strlen_data): New structure.
5540         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5541         For a type 2 call, if c_strlen indicates a non-terminated string
5542         use the length of the non-terminated string.
5543         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5545 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5547         PR target/87467
5548         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5549         __m512d type for __A argument rather than __m512.
5551 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5553         * match.pd (simple_comparison): Don't optimize if either operand is
5554         a function pointer when target needs function pointer canonicalization.
5556 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5558         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5559         power5 .. power9 to remove indirection.
5560         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5561         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5562         ASM_CPU_476_SPEC): Delete.
5563         (ASM_CPU_SPEC): Adjust.
5564         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5565         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5567 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5569         * config.in: Delete HAVE_AS_DCI.
5570         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5571         * config/rs6000/rs6000.h: Ditto.
5572         * configure.ac: Delete HAVE_AS_DCI.
5573         * configure: Regenerate.
5575 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5577         * config.in (HAVE_AS_LWSYNC): Delete.
5578         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5579         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5580         do it as a .long .
5581         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5582         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5583         as a .long .
5584         * configure.ac: Delete HAVE_AS_LWSYNC.
5585         * configure: Regenerate.
5587 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5588             Pierre-Marie de Rodat  <derodat@adacore.com>
5590         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5591         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5592         (cgraph_node::create_thunk): Add indirect_offset parameter.
5593         (thunk_adjust): Likewise.
5594         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5595         and initialize the corresponding field with it.
5596         (cgraph_node::dump): Dump indirect_offset field.
5597         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5598         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5599         (thunk_adjust): Add indirect_offset parameter and deal with it.
5600         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5601         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5602         or if the thunk is external or local.  Fix formatting.  Do not chain
5603         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5604         if any, in the GIMPLE representation.
5605         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5606         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5607         (input_node): Read indirect_offset field.
5608         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5609         call to thunk_adjust.
5610         * tree-nested.c (struct nesting_info): Add thunk_p field.
5611         (create_nesting_tree): Set it.
5612         (convert_all_function_calls): Copy static chain from targets to thunks.
5613         (finalize_nesting_tree_1): Return early for thunks.
5614         (unnest_nesting_tree_1): Do not finalize thunks.
5615         (gimplify_all_functions): Do not gimplify thunks.
5617 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5619         * opt-suggestions.c (option_proposer::build_option_suggestions):
5620         Release "option_values".
5622 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5624         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5625         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5626         * dumpfile.c (kind_as_string): New function.
5627         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5628         kind_as_string to vary the prefix based on dump_kind.
5629         (selftest::test_capture_of_dump_calls): Update for above.
5631 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5633         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5634         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5636 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5638         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5639         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5640         INVALID_REGNUM instead of FPSR_REG.
5641         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5642         * config/i386/i386.md: Update comment of FP compares.
5643         (fldenv): Do not clobber FPSR_REG.
5645 2018-09-28  Richard Biener  <rguenther@suse.de>
5647         * tree.h (BLOCK_ORIGIN): New.
5648         * omp-expand.c (grid_expand_target_grid_body): Assign
5649         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5650         * tree-inline.c (remap_block): Likewise.
5651         * auto-profile.c (get_function_decl_from_block): Simplify
5652         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5653         * langhooks.c (lhd_print_error_function): Likewise.
5654         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5655         Likewise.
5656         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5657         * tree.c (block_nonartificial_location): Likewise.
5658         (block_ultimate_origin): Likewise.
5659         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5660         no longer needed LTO case.
5662 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5663             Jan Hubicka  <jh@suse.cz>
5664             Martin Jambor  <mjambor@suse.cz>
5666         * simplify-rtx.c (simplify_merge_mask): New function.
5667         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5668         same masks are used in op1 or op2.
5669         (test_vec_merge): New function.
5670         (test_vector_ops): Call test_vec_merge.
5672 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5674         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5675         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5676         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5677         (cypress_costs): Set it.
5678         (supersparc_costs): Likewise.
5679         (hypersparc_costs): Likewise.
5680         (leon_cost): Likewise.
5681         (leon3_costs): Likewise.
5682         (sparclet_costs): Likewise.
5683         (ultrasparc_costs): Likewise.
5684         (ultrasparc_costs): Likewise.
5685         (niagara_costs): Likewise.
5686         (niagara2_costs): Likewise.
5687         (niagara3_costs): Likewise.
5688         (niagara4_costs): Likewise.
5689         (niagara7_costs): Likewise.
5690         (m8_costs): Likewise.
5691         (TARGET_CAN_FOLLOW_JUMP): Define.
5692         (pass_work_around_errata::gate): Minor tweak.
5693         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5694         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5695         Automaitcally clear MASK_FSMULD mask for V7 processors.
5696         (sparc_can_follow_jump): New static function.
5697         (output_ubranch): Deal with CROSSING_JUMP_P.
5698         (sparc_use_sched_lookahead): Rewrite using switch statement.
5699         (sparc_issue_rate): Reorder.
5700         (sparc_branch_cost): New function.
5702 2018-09-27  Martin Sebor  <msebor@redhat.com>
5704         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5705         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5706         (int_fits_type_p): Same.
5708 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5710         * config/i386/i386.md (FPCR_REG): Remove.
5711         (UNSPEC_FLDCW): Remove.
5712         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5713         (x86_fldcw_1): Remove insn pattern.
5714         (fnstenv): Do not clobber FPCR_REG.
5715         (fldenv): Ditto.
5716         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5717         (CALL_USED_REGISTERS): Ditto.
5718         (REG_ALLOC_ORDER): Ditto.
5719         (REG_CLASS_CONTENTS): Ditto.
5720         (HI_REGISTER_NAMES): Ditto.
5721         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5722         of numerical constants.
5723         * config/i386/i386.c (regclass_map): Remove fpsr register.
5724         (dbx_register_map): Ditto.
5725         (dbx64_register_map): Ditto.
5726         (svr4_dbx_register_map): Ditto.
5727         (print_reg): Do not handle FPCR_REG.
5729 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5731         PR target/87149
5732         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5733         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5734         Delete, always treat as true.
5735         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5736         Ditto.  Simplify remaining code.
5737         * config/powerpcspe/powerpcspe.h: Ditto.
5738         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5739         Simplify remaining code.
5740         (rs6000_expand_builtin): Ditto.
5741         * config/rs6000/rs6000.h: Ditto.
5742         * configure.ac: Ditto.
5743         * configure: Regenerate.
5745 2018-09-27  Martin Liska  <mliska@suse.cz>
5747         * coverage.c (get_coverage_counts): Revert the formatting
5748         of missing profile opt info.
5750 2018-09-27  Richard Biener  <rguenther@suse.de>
5752         PR debug/37801
5753         PR debug/87440
5754         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5755         block as we do not output that.
5756         (gen_inlined_subroutine_die): Elide the originally outermost
5757         block, matching what we do for concrete instances.
5758         (decls_for_scope): Add parameter specifying whether to recurse
5759         to subblocks.
5761 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5762             Tom de Vries  <tom@codesourcery.com>
5764         PR 82089
5766         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5767         STORE_FLAG_VALUE == 1.
5769 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5771         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5772         constant definitions.
5773         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5774         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5775         ("speculation_barrier"): New expander definition.
5777 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5779         PR gcov-profile/86957
5780         * common.opt: New warning option -Wmissing-profile.
5781         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5782         * doc/invoke.texi: Document -Wmissing-profile.
5784 2018-09-26  Jim Wilson  <jimw@sifive.com>
5786         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5787         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5789 2018-09-26  Martin Sebor  <msebor@redhat.com>
5791         * tree.c (zerop): Change return type to bool.
5792         (integer_zerop, integer_onep, integer_each_onep): Same.
5793         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5794         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5795         (real_onep, real_minus_onep, chain_index): Same.
5796         (print_type_hash_statistics, type_list_equal): Same.
5797         * tree.h (zerop): Same.
5798         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5799         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5800         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5801         (real_onep, real_minus_onep, chain_index): Same.
5802         (print_type_hash_statistics, type_list_equal): Same.
5804 2018-09-26  Jim Wilson  <jimw@sifive.com>
5806         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5808 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5810         PR target/87414
5811         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5812         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5813         call.
5815 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5817         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5819 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5821         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5822         and FP_SECOND_SSE_REGS.
5823         (REG_CLASS_NAMES): Ditto.
5824         (REG_CLASS_CONTENTS): Ditto.
5825         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5826         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5827         (ix86_preferred_output_reload_class): Ditto.
5828         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5829         clobber constraint to "=&f".
5830         (fix_truncdi_i387): Ditto.
5831         (lrintxfdi2): Ditto.
5832         (fistdi2_<rounding>): Ditto.
5833         (fpremxf4_i387): Change "=u" constraint to "=f".
5834         (fprem1xf4_i387): Ditto.
5835         (sincosxf3): Ditto.
5836         (fptanxf4_i387): Ditto.
5837         (fxtractxf3_i387): Ditto.
5838         (fscalexf4_i387): Ditto.
5839         (atan2xf3): Change "u" constraint to "f".
5840         (fyl2xxf3_i387): Ditto.
5841         (fyl2xp1xf3_i387): Ditto.
5843 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5845         PR target/87439
5846         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5847         for removed I387_MASK_PM entity.
5850 2018-09-26  Jeff Law  <law@redhat.com>
5851         Revert
5852         2018-09-26  Alexey Neyman  <stilor@att.net>
5854         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5855         headers are no longer pulled in by <isl/val.h>.
5857 2018-09-26  Richard Biener  <rguenther@suse.de>
5859         PR debug/87443
5860         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5861         or concrete instance DIE to the tree.  Create abstract origin
5862         attributes also for concrete instances.
5864 2018-09-26  Alexey Neyman  <stilor@att.net>
5866         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5867         headers are no longer pulled in by <isl/val.h>.
5869 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5871         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5872         Use new helper functions.
5873         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5874         Use new helper functions.
5875         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5876         aarch_mm_needs_release): New declarations.
5877         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5878         aarch_mm_needs_release): New.
5880 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
5882         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5883         (arm32_output_mi_thunk): Deal with long calls.
5885 2018-09-26  Richard Biener  <rguenther@suse.de>
5887         PR debug/87428
5888         PR debug/87362
5889         * tree-inline.c (expand_call_inline): When the location
5890         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5891         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5892         the inserted BLOCK to make inlined_function_outer_scope_p
5893         recognize it.
5894         * dwarf2out.c (add_call_src_coords_attributes): Do not add
5895         coords for reserved locations.
5897 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5899         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5900         (*call_indirect_nonlocal_sysv<mode>): Ditto.
5901         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5902         (*sibcall_nonlocal_sysv<mode>): Ditto.
5903         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5904         (<bd>_<mode>): Ditto.
5905         (<bd>tf_<mode>): Ditto.
5907 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5909         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5910         control string as a list of templates instead of as C code.
5911         (*altivec_movti): Ditto.
5912         * config/rs6000/darwin.md (movdf_low_di): Ditto.
5914 2018-09-25  Jim Wilson  <jimw@sifive.com>
5916         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5917         when target symbol is weak.
5919 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5921         PR c/87387
5922         * builtins.c (unterminated_array): Simplify.
5923         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
5924         where pointer arithmetic is safe.
5926 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
5928         PR target/86987
5929         * config/rs6000/altivec.md (altivec_vspltb): Use
5930         const_0_to_15_operand instead of u5bit_cint_operand.
5931         (*altivec_vspltb_internal): Ditto.
5932         (altivec_vspltb_direct): Ditto.
5933         (altivec_vsplth): Use const_0_to_7_operand instead of
5934         u5bit_cint_operand.
5935         (*altivec_vsplth_internal): Ditto.
5936         (altivec_vsplth_direct): Ditto.
5937         (altivec_vspltw): Use const_0_to_3_operand instead of
5938         u5bit_cint_operand.
5939         (*altivec_vspltw_internal): Ditto.
5940         (altivec_vspltw_direct): Ditto.
5941         (altivec_vspltsf): Ditto.
5942         (*altivec_vspltsf_internal): Ditto.
5943         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5944         various splats with the proper size immediate.  Reorder the various
5945         cases by ascending size of immediate, and put all such together.
5947 2018-09-25  Richard Biener  <rguenther@suse.de>
5949         PR debug/83941
5950         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5951         GC-ification.
5952         (maybe_create_die_with_external_ref): Do not create
5953         DW_TAG_imported_unit here.
5954         (add_abstract_origin_attribute): Handle external BLOCK refs.
5955         (dwarf2out_abstract_function): Simplify LTO case.
5956         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5957         rather than using maybe_create_die_with_external_ref.
5959 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
5961         PR target/71278
5962         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5963         (frndintxf2_mask_pm_i387): Ditto.
5964         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5965         Enable for !flag_trapping_math.
5966         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5967         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5968         Change operand 1 predicate to nonimmediate_operand.
5969         (attr "i387_cw"): Remove mask_pm.
5970         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5971         (enum ix86_entity): Remove I387_MASK_PM.
5972         * config/i386/i386.c (ix86_i387_mode_needed): Do not
5973         handle I387_MASK_PM.
5974         (ix86_mode_needed): Ditto.
5975         (ix86_mode_after): Ditto.
5976         (ix86_mode_entry): Ditto.
5977         (ix86_mode_exit): Ditto.
5978         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5980 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
5982         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5983         to_update_switch_stmts to vNULL instead of calling create on them
5984         immediately.
5986 2018-09-25  Richard Biener  <rguenther@suse.de>
5988         PR tree-optimization/87402
5989         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5990         (visit_phi): Re-instantiate handling of supposed to be VARYING
5991         but non-VARYING backedge value.
5993 2018-09-25  Richard Biener  <rguenther@suse.de>
5995         PR debug/83941
5996         * dwarf2out.c (struct sym_off_pair): New.
5997         (external_die_map): New global.
5998         (lookup_decl_die): When in LTO create DIEs lazily from the
5999         external_die_map.
6000         (lookup_block_die): New function, create DIEs lazily in LTO.
6001         (equate_block_to_die): New function.
6002         (dwarf2out_die_ref_for_decl): During WPA get the association
6003         from the external DIE map.
6004         (dwarf2out_register_external_die): Record mapping into the
6005         external DIE map.
6006         (maybe_create_die_with_external_ref): New function split out from
6007         DIE generation part of old dwarf2out_register_external_die.
6008         (add_abstract_origin_attribute): Do not return the DIE.  When
6009         in LTO reference externals directly.
6010         (dwarf2out_abstract_function): When in LTO ignore calls for
6011         decls with external DIEs (already present abstract instances).
6012         (gen_call_site_die): Adjust.
6013         (add_high_low_attributes): Likewise.
6014         (gen_lexical_block_die): Likewise.
6015         (gen_inlined_subroutine_die): Likewie.
6016         (gen_block_die): Likewise.
6017         (dwarf2out_inline_entry): Likewise.
6018         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6019         DIEs.
6021 2018-09-25  Martin Liska  <mliska@suse.cz>
6023         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6024         integers and not by a float value.
6026 2018-09-25  Martin Liska  <mliska@suse.cz>
6028         PR fortran/87394
6029         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6030         instead of NULL.
6031         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6032         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6033         * godump.c (go_format_type): Remove extra parenthesis.
6035 2018-09-25  Martin Liska  <mliska@suse.cz>
6037         * alias.c (set_dest_equal_p): Remove unused function.
6038         * config/i386/i386.c (def_builtin_pure2): Likewise.
6039         * diagnostic-show-locus.c (class layout): Remove
6040         unused field.
6041         (layout::layout): Likewise here.
6042         * dump-context.h (class temp_dump_context): Likewise.
6043         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6044         (add_AT_loclistsptr): Likewise.
6045         (add_AT_offset): Likewise.
6046         (get_AT_hi_pc): Likewise.
6047         (is_comdat_die): Likewise.
6048         (type_is_enum): Likewise.
6049         (ceiling): Likewise.
6050         (add_AT_vms_delta): Likewise.
6051         (is_class_die): Likewise.
6052         * edit-context.c (class line_event): Remove unused field.
6053         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6054         unused function.
6055         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6056         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6057         (ok_for_base_p_nonstrict): Likewise.
6058         * tree-chrec.c (is_not_constant_evolution): Likewise.
6059         (chrec_fold_poly_cst): Likewise.
6060         * tree-if-conv.c (has_pred_critical_p): Likewise.
6061         * tree-ssa-coalesce.c (print_exprs): Likewise.
6062         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6063         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6064         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6065         (value_range_nonnegative_p): Likewise.
6067 2018-09-25  Martin Liska  <mliska@suse.cz>
6069         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6070         Do not handle "GNU Pascal".
6071         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6072         Likewise.
6073         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6074         from documentation. Likewise.
6075         * dbxout.c (dbxout_range_type): Likewise.
6076         * doc/cpp.texi: Likewise.
6077         * doc/extend.texi: Likewise.
6078         * doc/frontends.texi: Likewise.
6079         * doc/invoke.texi: Remove Pascal entry.
6080         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6081         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6083 2018-09-25  Martin Liska  <mliska@suse.cz>
6085         PR middle-end/86078
6086         * doc/invoke.texi: Document all parameters and remove default
6087         of the parameters.
6089 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6091         PR bootstrap/87417
6092         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6093         contains HOST_WIDE_INTs when computing its size.
6095 2018-09-24  Jim Wilson  <jimw@sifive.com>
6097         PR target/87391
6098         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6099         not TARGET_RVE.
6100         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6102 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6104         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6105         access prev before checking it for NULLness in the
6106         AARCH64_FUSE_CMP_BRANCH case.
6108 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6110         PR target/82699
6111         * config/i386/i386.c (rest_of_insert_endbranch): Set
6112         endbr_queued_at_entrance to true and don't insert ENDBR if
6113         x86_function_profiler will be called.
6114         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6115         is true.
6116         * config/i386/i386.h (machine_function): Add
6117         endbr_queued_at_entrance.
6119 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6121         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6122         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6123         types.
6124         (check_attr_test): Use alternative_mask.
6125         (get_attr_value): Likewise.
6126         (compute_alternative_mask): Use alternative_mask and XWINT.
6127         (make_alternative_compare): Use alternative_mask.
6128         (attr_alt_subset_p): Use XWINT.
6129         (attr_alt_subset_of_compl_p): Likewise.
6130         (attr_alt_intersection): Use alternative_mask and XWINT.
6131         (attr_alt_union): Likewise.
6132         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6133         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6134         (simplify_test_exp): Use alternative_mask and XWINT.
6135         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6136         number calculation to support 64 bits.  Generate code that
6137         checks 64-bit masks.
6138         (main): Use alternative_mask.
6139         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6141 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6143         PR target/80080
6144         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6145         RETURN+USE when returning via %r14.
6147 2018-09-24  Martin Liska  <mliska@suse.cz>
6149         * gcov.c (output_lines): Print colorization legend
6150         for both flag_use_colors and flag_use_hotness_colors.
6151         Reword the help.
6153 2018-09-24  Martin Liska  <mliska@suse.cz>
6155         * coverage.c (get_coverage_counts): Use warning_at
6156         with current_function_decl location. Use %qD in warning
6157         message.
6159 2018-09-24  Martin Liska  <mliska@suse.cz>
6161         * memory-block.h (memory_block_pool::release): Annotate with
6162         valgrind that the memory is not accessible.
6164 2018-09-24  Martin Liska  <mliska@suse.cz>
6166         PR sanitizer/85774
6167         * asan.c: Make asan_handled_variables extern.
6168         * asan.h: Likewise.
6169         * cfgexpand.c (expand_stack_vars): Make sure
6170         a representative is unpoison if another
6171         variable in the partition is handled by
6172         use-after-scope sanitization.
6174 2018-09-24  Richard Biener  <rguenther@suse.de>
6176         PR tree-optimization/63155
6177         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6178         the worklist when the edge of the respective argument isn't
6179         executable.
6181 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6183         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6184         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6185         (MASK_CLASS_P): Update for rename.
6186         (MAYBE_MASK_CLASS_P): Ditto.
6187         (REG_CLASS_NAMES): Update.
6188         (REG_CLASS_CONTENT): Update.
6189         * config/i386/i386.c (regclass_map): Update for MASK_REG
6190         and ALL_MASK_REGS rename.
6191         * config/i386/constraints.md (Yk): Update for rename.
6192         (k): Ditto.
6194 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6196         * config/i386/i386.h (enum reg_class): Remove
6197         EVEX_SSE_REGS and MOD4_SSE_REGS.
6198         (REG_CLASS_NAMES): Update.
6199         (REG_CLASS_CONTENT): Update.
6200         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6201         registers as ALL_SSE_REGS.
6202         (ix86_additional_allocno_class_p): Remove.
6203         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6204         (ix86_register_priority): Lower priority of EVEX SSE registers.
6205         Use IN_RANGE macro where appropriate.
6206         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6207         AVX-5124VNNIW checks.
6208         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6209         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6210         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6211         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6212         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6213         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6214         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6215         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6216         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6217         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6218         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6219         * config/i386/constraints.md (Yh): Remove.
6221 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6223         * config/i386/i386.c (regclass_map): Declare integer REX registers
6224         as GENERAL_REGS.
6226 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6228         * doc/service.texi (Service): Switch the fsf.org link to https.
6230 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6232         PR target/86798
6233         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6234         Define to speculation_safe_value_not_needed.
6236 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6238         PR middle-end/81035
6239         * doc/extend.texi (Common Function Attributes): Mention that
6240         noreturn suppresses tail call optimization.
6242 2018-09-21  Jeff Law  <law@redhat.com>
6244         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6245         vr_values::cleanup_edges_and_switches.
6246         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6247         vr_values class.
6248         (identify_jump_threads): Remove EDGE_IGNORE handling.
6249         (execute_vrp): Move handling of to_remove_edges and
6250         to_update_switch_stmts into vr_values class member functions.
6251         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6252         (to_update_switch_stmts): Likewise.
6253         * vr-values.c: Include cfghooks.h.
6254         (vr_values::vr_values): Initialize to_remove_edges and
6255         to_update_switch_stmts.
6256         (vr_values::~vr_values): Verify to_remove_edges and
6257         to_update_switch_stmts are empty.
6258         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6259         (vr_values::cleanup_edges_and_switches): New member function.
6260         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6261         function.  Add new data members.
6263 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6265         PR tree-optimization/87309
6266         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6267         calls with pflags and alt_flags.
6268         (selftest::test_capture_of_dump_calls): Add test of interaction of
6269         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6271 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6273         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6275 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6277         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6279 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6281         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6282         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6283         (VXWORKS_LIBS_RTP): Minor reordering.
6285 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6287         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6288         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6290 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6292         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6293         (PTRDIFF_TYPE): Likewise.
6295 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6297         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6298         triplet, similar to support for VxWorks7.
6299         * config/vxworks-dummy.h: Provide a default definition
6300         of TARGET_VXWORKS64 to 0.
6302 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6304         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6305         * config/vxworks-dummy.h: here.
6307 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6309         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6311 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6312             Bo Zhou  <zbo.zhou@hisilicon.com>
6314         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6315         * config/aarch64/aarch64-tune.md: Regenerated.
6316         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6317         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6318         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6320 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6321             Julian Brown  <julian@codesourcery.com>
6323         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6325 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6327         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6328         if the call takes a static chain.
6330 2018-09-21  Martin Liska  <mliska@suse.cz>
6332         * auto-profile.c (autofdo_source_profile::read): Do not
6333         set sum_all.
6334         (read_profile): Do not add working sets.
6335         (read_autofdo_file): Remove sum_all.
6336         (afdo_callsite_hot_enough_for_early_inline): Remove const
6337         qualifier.
6338         * coverage.c (struct counts_entry): Remove gcov_summary.
6339         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6340         do not support GCOV_TAG_PROGRAM_SUMMARY.
6341         (get_coverage_counts): Remove summary and expected
6342         arguments.
6343         * coverage.h (get_coverage_counts): Likewise.
6344         * doc/gcov-dump.texi: Remove -w option.
6345         * gcov-dump.c (dump_working_sets): Remove.
6346         (main): Do not support '-w' option.
6347         (print_usage): Likewise.
6348         (tag_summary): Likewise.
6349         * gcov-io.c (gcov_write_summary): Do not dump
6350         histogram.
6351         (gcov_read_summary): Likewise.
6352         (gcov_histo_index): Remove.
6353         (gcov_histogram_merge): Likewise.
6354         (compute_working_sets): Likewise.
6355         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6356         it not obsolete.
6357         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6358         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6359         (GCOV_HISTOGRAM_SIZE): Remove.
6360         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6361         (struct gcov_summary): Simplify rapidly just
6362         to runs and sum_max fields.
6363         (gcov_histo_index): Remove.
6364         (NUM_GCOV_WORKING_SETS): Likewise.
6365         (compute_working_sets): Likewise.
6366         * gcov-tool.c (print_overlap_usage_message): Remove
6367         trailing empty line.
6368         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6369         (output_lines): Remove program related line.
6370         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6371         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6372         histogram.
6373         (input_profile_summary): Do not read it.
6374         (merge_profile_summaries): And do not merge it.
6375         (input_symtab): Do not call removed function.
6376         * modulo-sched.c (sms_schedule): Do not print sum_max.
6377         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6378         removed when histogram method was invented.
6379         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6380         mode.
6381         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6382         GCOV coding style.
6383         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6384         and dump selected value.
6385         * profile.c (add_working_set): Remove.
6386         (get_working_sets): Likewise.
6387         (find_working_set): Likewise.
6388         (get_exec_counts): Do not work with working sets.
6389         (read_profile_edge_counts): Do not inform as sum_max is removed.
6390         (compute_branch_probabilities): Likewise.
6391         (compute_value_histograms): Remove argument for call of
6392         get_coverage_counts.
6393         * profile.h: Do not make gcov_summary const.
6395 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6397         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6399 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6401         PR tree-optimization/86990
6402         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6403         Check that the entire merged store group is made of constants only for
6404         overlapping stores.
6406 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6408         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6409         (VTABLE_VERIFICATION_SPEC): Likewise.
6410         (SANITIZER_EARLY_SPEC): Likewise.
6411         (SANITIZER_SPEC): Likewise.
6412         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6413         * doc/invoke.texi (Link Options): Document -r.
6415 2018-09-20  Richard Biener <rguenther@suse.de>
6417         PR middle-end/87054
6418         * gimplify.c (gimplify_expr): Retain alignment of
6419         addressable lvalue in dereference.
6421 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6423         PR bootstrap/87013
6424         * configure.ac: Check for .loc is_stmt support.
6425         * configure, config.in: Rebuilt.
6426         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6427         if not supported.
6429 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6431         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6432         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6433         -misel=no.
6435 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6437         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6438         VECTOR_OTHER.
6439         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6440         case VECTOR_OTHER.
6442 2018-09-20  Marek Polacek  <polacek@redhat.com>
6444         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6446 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6448         PR tree-optimization/87288
6449         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6450         into account when determining PEELING_FOR_NITERS.
6452 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6454         PR tree-optimization/86877
6455         * tree-vect-loop.c (vect_analyze_loop_2): Call
6456         vect_verify_datarefs_alignment.
6458 2018-09-19  Marek Polacek  <polacek@redhat.com>
6460         * doc/invoke.texi: Document -Wclass-conversion.
6462 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6464         * config/pa/pa.c (pa_adjust_priority): Delete.
6465         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6467         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6468         (atomic_storehi): Likewise.
6469         (atomic_storesi): Likewise.
6470         (atomic_loaddi): Restore compare and swap exchange loop code.
6472 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6474         PR rtl-optimization/86902
6475         * combine.c (try_combine): When changing the CC mode used, don't change
6476         an unrelated mode in other_insn to that new CC mode.
6478 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6480         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6481         with %T in place of calls to dump_generic_expr.
6482         (prune_runtime_alias_test_list): Likewise.
6483         (create_runtime_alias_checks): Likewise.
6484         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6485         (vect_analyze_data_ref_dependence): Likewise.
6486         (vect_slp_analyze_data_ref_dependence): Likewise.
6487         (vect_record_base_alignment): Likewise.  Use %G in place of call
6488         to dump_gimple_stmt.
6489         (vect_compute_data_ref_alignment): Likewise.
6490         (verify_data_ref_alignment): Likewise.
6491         (vect_find_same_alignment_drs): Likewise.
6492         (vect_analyze_group_access_1): Likewise.
6493         (vect_analyze_data_ref_accesses): Likewise.
6494         (dependence_distance_ge_vf): Likewise.
6495         (dump_lower_bound): Likewise.
6496         (vect_prune_runtime_alias_test_list): Likewise.
6497         (vect_find_stmt_data_reference): Likewise.
6498         (vect_analyze_data_refs): Likewise.
6499         (vect_create_addr_base_for_vector_ref): Likewise.
6500         (vect_create_data_ref_ptr): Likewise.
6501         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6502         (vect_can_advance_ivs_p): Likewise.
6503         (vect_update_ivs_after_vectorizer): Likewise.
6504         (vect_gen_prolog_loop_niters): Likewise.
6505         (vect_prepare_for_masked_peels): Likewise.
6506         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6507         (vect_determine_vectorization_factor): Likewise.
6508         (vect_is_simple_iv_evolution): Likewise.
6509         (vect_analyze_scalar_cycles_1): Likewise.
6510         (vect_analyze_loop_operations): Likewise.
6511         (report_vect_op): Likewise.
6512         (vect_is_slp_reduction): Likewise.
6513         (check_reduction_path): Likewise.
6514         (vect_is_simple_reduction): Likewise.
6515         (vect_create_epilog_for_reduction): Likewise.
6516         (vect_finalize_reduction:): Likewise.
6517         (vectorizable_induction): Likewise.
6518         (vect_transform_loop_stmt): Likewise.
6519         (vect_transform_loop): Likewise.
6520         (optimize_mask_stores): Likewise.
6521         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6522         (vect_split_statement): Likewise.
6523         (vect_recog_over_widening_pattern): Likewise.
6524         (vect_recog_average_pattern): Likewise.
6525         (vect_determine_min_output_precision_1): Likewise.
6526         (vect_determine_precisions_from_range): Likewise.
6527         (vect_determine_precisions_from_users): Likewise.
6528         (vect_mark_pattern_stmts): Likewise.
6529         (vect_pattern_recog_1): Likewise.
6530         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6531         (vect_record_max_nunits): Likewise.
6532         (vect_build_slp_tree_1): Likewise.
6533         (vect_build_slp_tree_2): Likewise.
6534         (vect_print_slp_tree): Likewise.
6535         (vect_analyze_slp_instance): Likewise.
6536         (vect_detect_hybrid_slp_stmts): Likewise.
6537         (vect_detect_hybrid_slp_1): Likewise.
6538         (vect_slp_analyze_operations): Likewise.
6539         (vect_slp_analyze_bb_1): Likewise.
6540         (vect_transform_slp_perm_load): Likewise.
6541         (vect_schedule_slp_instance): Likewise.
6542         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6543         (vect_mark_stmts_to_be_vectorized): Likewise.
6544         (vect_init_vector_1): Likewise.
6545         (vect_get_vec_def_for_operand): Likewise.
6546         (vect_finish_stmt_generation_1): Likewise.
6547         (vect_check_load_store_mask): Likewise.
6548         (vectorizable_call): Likewise.
6549         (vectorizable_conversion): Likewise.
6550         (vectorizable_operation): Likewise.
6551         (vectorizable_load): Likewise.
6552         (vect_analyze_stmt): Likewise.
6553         (vect_is_simple_use): Likewise.
6554         (vect_get_vector_types_for_stmt): Likewise.
6555         (vect_get_mask_type_for_stmt): Likewise.
6556         * tree-vectorizer.c (increase_alignment): Likewise.
6558 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6560         * doc/rtl.texi: Adjust vec_select description.
6561         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6562         non-constant selectors.
6564 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6566         * config/aarch64/aarch64-protos.h
6567         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6568         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6569         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6570         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6571         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6572         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6573         (AARCH64_ISA_RCPC8_4): New flag.
6574         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6575         to aarch64_offset_9bit_signed_unscaled_p.
6576         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6577         and use stlur.
6578         * config/aarch64/constraints.md (Ust): New constraint.
6579         * config/aarch64/predicates.md.
6580         (aarch64_9bit_offset_memory_operand): New predicate.
6581         (aarch64_rcpc_memory_operand): New predicate.
6583 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6585         PR rtl-optimization/87361
6586         * rtlanal.c (nonzero_bits1): Revert accidental change.
6588 2018-09-19  Richard Biener  <rguenther@suse.de>
6590         PR tree-optimization/87349
6591         PR tree-optimization/87342
6592         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6594 2018-09-18  Marek Polacek  <polacek@redhat.com>
6596         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6597         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6599 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6601         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6603 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6605         PR rtl-optimization/86882
6606         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6608 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6610         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6611         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6613 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6615         PR other/87353
6616         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6618 2018-09-18  Richard Biener  <rguenther@suse.de>
6620         PR middle-end/63155
6621         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6622         (compute_samebase_partition_bases): Likewise.
6623         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6624         (gimple_can_coalesce_p): Simplify.
6626 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6628         Handle a library implementation of ffs calling __builtin_ffs.
6629         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6630         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6631         to __ffsdi2.
6633 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6635         * diagnostic-show-locus.c (class layout_range): Add field
6636         "m_original_idx".
6637         (layout_range::layout_range): Add "original_idx" param and use it
6638         to initialize new field.
6639         (make_range): Use 0 for original_idx.
6640         (layout::layout): Pass in index to calls to
6641         maybe_add_location_range.
6642         (layout::maybe_add_location_range): Add param "original_idx" and
6643         pass it on to layout_range.
6644         (layout::print_any_labels): Pass on range->m_original_idx to
6645         get_text call.
6646         (gcc_rich_location::add_location_if_nearby): Use 0 for
6647         original_idx.
6648         * gcc-rich-location.h (text_range_label::get_text): Update for new
6649         param.
6650         (range_label_for_type_mismatch::get_text): Likewise.
6652 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6654         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6656 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6658         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6659         format_string_diagnostic_t.
6660         (fmtwarn_n): Likewise.
6661         * substring-locations.c
6662         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6663         (format_warning_n_va): Convert to...
6664         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6665         (format_warning_va): Convert to...
6666         (format_string_diagnostic_t::emit_warning_va): ...this.
6667         (format_warning_at_substring): Convert to...
6668         (format_string_diagnostic_t::emit_warning): ...this.
6669         (format_warning_at_substring_n): Convert to...
6670         (format_string_diagnostic_t::emit_warning_n): ...this.
6671         * substring-locations.h (class format_string_diagnostic_t): New
6672         class.
6673         (format_warning_va): Convert to
6674         format_string_diagnostic_t::emit_warning_va.
6675         (format_warning_n_va): Convert to
6676         format_string_diagnostic_t::emit_warning_n_va.
6677         (format_warning_at_substring): Convert to
6678         format_string_diagnostic_t::emit_warning.
6679         (format_warning_at_substring_n): Convert to
6680         format_string_diagnostic_t::emit_warning_n.
6682 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6683             Bernd Schmidt <bernds_cb1@t-online.de>
6685         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6686         SImode args.
6688 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6690         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6691         operand 0 predicate to nonimmediate operand.
6692         (rint<mode>2_frndint): Remove insn pattern.
6693         (rint<mode>2): Change operand 1 predicate to general_operand.
6694         Extend operand 1 to XFmode and generate rintxf2 insn.
6695         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6696         Do not use X87MODEF mode macro.
6697         (frndintxf2_<rounding>_i387): Rename from
6698         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6699         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6700         to XFmode and generate significandxf3 insn.
6702 2018-09-17  Richard Biener  <rguenther@suse.de>
6704         PR tree-optimization/87328
6705         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6706         visiting unexecutable backedges when not iterating.
6707         (do_rpo_vn): Mark all edges not executable even when not
6708         iterating.
6710 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6712         PR c/63886
6713         * doc/invoke.texi (Warning Options): Likewise.
6715 2018-09-17  Richard Biener  <rguenther@suse.de>
6717         PR tree-optimization/87301
6718         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6719         clean EH info from leftover copy assignments.
6721 2018-09-17  Martin Liska  <mliska@suse.cz>
6723         PR gcov-profile/85871
6724         * gcov.c (output_intermediate_file): Fix out of bounds
6725         access.
6727 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6729         * config/arc/arc.c: Object attributes for core4 not reflected
6730         correctly.
6731         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6732         core3).
6734 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6736         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6738 2018-09-17  Martin Liska  <mliska@suse.cz>
6740         * doc/gcov.texi: Document new option --use-hotness-colors.
6741         * gcov.c (struct source_info): Declare new field.
6742         (source_info::source_info): Set default for maximum_count.
6743         (print_usage): Add new -q option.
6744         (process_args): Process it.
6745         (accumulate_line_info): Save src->maximum_count.
6746         (output_line_beginning): Make color line number if
6747         flag_use_hotness_colors is set.
6748         (output_line_details): Pass default argument value.
6749         (output_lines): Pass src->maximum_count.
6751 2018-09-17  Martin Liska  <mliska@suse.cz>
6753         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6754         Use processor_names table.
6755         * config/i386/i386.c (ix86_default_align): Use
6756         processor_cost_table for alignment values.
6757         (ix86_option_override_internal): Use processor_names.
6758         (ix86_function_specific_print): Likewise.
6759         * config/i386/i386.h (struct processor_costs):
6760         Add alignment values.
6761         (struct ptt): Remove and replace with const char *.
6762         * config/i386/x86-tune-costs.h (struct processor_costs):
6763         Declare default alignments for all costs.
6765 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6767         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6768         symbolics or VR_VARYING ranges for ABS_EXPR.
6769         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6770         when range will wrap.
6772 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6774         PR middle-end/86864
6775         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6776         before and after a JUMP_TABLE_DATA.
6778 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6780         PR middle-end/87188
6781         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6782         when one operand is a function pointer.  Use POINTER_TYPE_P and
6783         FUNC_OR_METHOD_TYPE_P.
6784         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6785         * fold-const.c (build_range_check): Likewise.
6786         * match.pd (simple_comparison): Likewise.
6788 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6790         PR c/82967
6791         * spellcheck.c (get_edit_distance_cutoff): New function.
6792         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6793         (selftest::test_get_edit_distance_one_way): ...this.
6794         (selftest::test_get_edit_distance_unit): Rename to...
6795         (selftest::test_get_edit_distance_both_ways): ...this.
6796         (selftest::test_edit_distances): Move tests to this new function,
6797         and test some more pairs of strings.  Update for above renaming.
6798         (selftest::get_old_cutoff): New function.
6799         (selftest::test_get_edit_distance_cutoff): New function.
6800         (selftest::assert_suggested_for): New function.
6801         (ASSERT_SUGGESTED_FOR): New macro.
6802         (selftest::assert_not_suggested_for): New function.
6803         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6804         (selftest::test_suggestions): New function.
6805         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6806         tests to selftest::test_edit_distances and call it.  Add calls to
6807         selftest::test_get_edit_distance_cutoff and
6808         selftest::test_suggestions.
6809         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6810         (best_match::consider): Replace hard-coded cutoff calculation with
6811         a call to...
6812         (best_match::get_cutoff): New declaration.
6813         (best_match::get_best_meaningful_candidate): Likewise.
6815 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6817         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6819 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6821         revert:
6822         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6824         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6825         terminated string literal.
6827 2018-09-14  Martin Sebor  <msebor@redhat.com>
6829         * builtins.c (unterminated_array): Handle ARRAY_REF.
6830         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6831         * builtins.h (unterminated_array): Declare extern.
6832         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6833         arrays.
6834         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6835         calls.
6837 2018-09-14  Martin Sebor  <msebor@redhat.com>
6838             Jeff Law  <law@redhat.com>
6840         * builtins.c (unterminated_array): New.
6841         (expand_builtin_strcpy): Adjust.
6842         (expand_builtin_strcpy_args): Detect unterminated arrays.
6843         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6844         unterminated arrays.
6845         * gimple-fold.h (get_maxval_strlen): Add argument.
6846         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6848         * gimple-fold.c (get_range_strlen): Add argument.
6849         (get_maxval_strlen): Adjust.
6850         * gimple-fold.h (get_range_strlen): Add argument.
6852 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6854         * config/i386/movdirintrin.h: Fix copyright year.
6856 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6858         * reg-stack.c: Include regs.h.
6859         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6860         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6861         FIRST_STACK_REG, not DFmode.
6862         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6863         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6864         (change stack): Default register mode to the reg_raw_mode of
6865         FIRST_STACK_REG, not DFmode.
6866         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6867         (*swapxf): Rename from swapxf.
6869 2018-09-14  Carl Love  <cel@us.ibm.com>
6871         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6872         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6874 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
6876         PR target/87224
6877         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6878         alternatives.
6880 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
6882         PR target/85628
6883         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6885 2018-09-14  Jason Merrill  <jason@redhat.com>
6887         Fix --enable-gather-detailed-mem-stats.
6888         * hash-table.c (hash_table_usage): Change from variable to function.
6889         * hash-table.h: Adjust.
6890         * Makefile.in: Add missing dependencies on hash-table.h.
6892 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6894         PR tree-optimization/87259
6895         PR lto/87283
6896         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6897         execute_cse_reciprocals_1 has tried transforming.
6899 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
6901         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6902         VR_VARYING for PLUS/MINUS_EXPR.
6904 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6906         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6907         formatting.
6909 2018-09-14  Richard Biener  <rguenther@suse.de>
6911         PR middle-end/63155
6912         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6913         bits for the merged partition.
6915 2018-09-13  Martin Sebor  <msebor@redhat.com>
6916             Bernd Edlinger  <bernd.edlinger@hotmail.de>
6918         * builtins.h (c_srlen): Add argument.
6919         * builtins.c (warn_string_no_nul): New function.
6920         (c_strlen): Add argument and use it.  Update recursive calls.
6921         Pass DECL argument to string_constant to get info on non
6922         terminated strings.  Update *NONSTR as needed.
6923         (fold_builtin_strlen): Add argument to calls to c_strlen.
6924         Warn for unterminated arrays.
6925         (warn_string_no_null): Add prototype.
6926         * expr.c (string_constant): Update arguments.  Update recursive
6927         calls appropriately.  Detect missing NUL terminator and outermost
6928         declaration its missing in.
6929         Improve checks for arrays with nonzero lower bound or elements
6930         that are not a single byte.  Simplify offset computation.
6931         Simplify checks for non-NUL terminated strings.
6932         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6933         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6935 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6937         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6938         correctly.
6939         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
6940         argument.  Fix range checks.
6941         * fold-const.h (c_getstr): Adjust protoype.
6942         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6943         string is constant but contains no NUL byte.
6945         * expr.c (string_constant): Adjust function comment.
6946         Remove bogus check for zero termination.
6948         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6950         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6951         (get_constant_size): Don't make STRING_CSTs larger than they are.
6952         (check_string_literal): New check function for STRING_CSTs.
6953         (output_constant): Use it.
6955 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
6957         PR target/86812
6958         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6960 2018-09-13  Richard Biener  <rguenther@suse.de>
6962         PR tree-optimization/87263
6963         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6964         (struct unwind_state): Add max_rpo field.
6965         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6966         Compute max_rpo, the max RPO number a block can be backwards reached
6967         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
6968         separating it from the iterating mode.
6970 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
6972         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6973         (rfs_decision): New scheduling decision.
6975 2018-09-13  Richard Biener  <rguenther@suse.de>
6977         PR bootstrap/87134
6978         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6979         (vn_nary_op_insert_pieces_predicated): Do not write useless
6980         valid_dominated_by_p entry outside of the allocated storage.
6982 2018-09-13  Omar Sandoval  <osandov@osandov.com>
6983             Tom de Vries  <tdevries@suse.de>
6985         PR debug/86985
6986         * dwarf2out.c (is_c): New function.
6987         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6989 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
6991         PR target/85628
6992         * config/aarch64/aarch64.md (*aarch64_bfxil):
6993         Define.
6994         * config/aarch64/constraints.md (Ulc): Define.
6995         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6996         Define.
6997         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6998         New function.
7000 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7002         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7003         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7004         aarch64_layout_frame call.
7005         (aarch64_expand_epilogue): Likewise.
7006         (aarch64_initial_elimination_offset): Likewise.
7007         (aarch64_get_separate_components): Likewise.
7008         (aarch64_use_return_insn_p): Likewise.
7009         (aarch64_layout_frame): Remove unneeded check.
7011 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7013         * configure.ac: Only append
7014         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7015         gcc_config_arguments if it was never reconfigured or last reconfigure
7016         was with different arguments.
7017         * configure: Regenerated.
7019 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7020             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7022         PR middle-end/87290
7023         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7024         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7026 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7028         PR tree-optimization/87287
7029         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7030         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7031         * match.pd (X % C == 0): ... here.  New optimization.
7033 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7035         PR middle-end/82853
7036         * expr.h (maybe_optimize_mod_cmp): Declare.
7037         * expr.c (mod_inv): New function.
7038         (maybe_optimize_mod_cmp): New function.
7039         (do_store_flag): Use it.
7040         * cfgexpand.c (expand_gimple_cond): Likewise.
7042 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7043             Julian Brown  <julian@codesourcery.com>
7045         PR middle-end/86336
7046         * gimplify.c (gimplify_scan_omp_clauses): Set
7047         target_firstprivatize_array_bases in OpenACC parallel and kernels
7048         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7049         OpenACC data regions.
7051 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7053         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7054         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7056 2018-09-12  Richard Biener  <rguenther@suse.de>
7058         PR tree-optimization/87280
7059         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7060         edge but unreachable target.
7061         (do_rpo_vn): For conservatively handling a PHI only mark
7062         the backedge executable but not the block reachable.
7064 2018-09-12  Richard Biener  <rguenther@suse.de>
7066         PR tree-optimization/87266
7067         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7068         visited blocks.
7070 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7072         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7073         constants.
7074         ("trunc<BFP:mode><DFP_ALL:mode>2")
7075         ("trunc<DFP_ALL:mode><BFP:mode>2")
7076         ("extend<BFP:mode><DFP_ALL:mode>2")
7077         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7078         according to the target operand type.
7080 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7081             Andreas Krebbel  <krebbel@linux.ibm.com>
7083         PR tree-optimization/86844
7084         * gimple-ssa-store-merging.c
7085         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7086         there are any overlapping stores in between them, make sure they are
7087         also coalesced or we give up completely.
7089 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7091         PR middle-end/87248
7092         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7093         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7095 2018-09-12  Tom de Vries  <tdevries@suse.de>
7097         * common.opt (gdescribe-dies): Add option.
7098         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7099         attribute for artifical and nameless decls.
7100         (dwarf2out_register_external_die): Add description attribute to
7101         external reference die.
7102         (add_desc_attribute): New functions.
7103         (gen_subprogram_die): Add description attribute to
7104         DW_TAG_call_site_parameter.
7105         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7106         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7107         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7108         -gno-describe-dies.
7109         (@item -gdescribe-dies): Add.
7111 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7113         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7114         (extract_range_from_binary_expr_1: Call
7115         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7116         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7117         depend on sign.
7119 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7121         * gimple-ssa-warn-alloca.c
7122         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7123         field for ALLOCA_BOUND_*_LARGE.
7125 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7127         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7129 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7131         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7132         for clobbers.  Remove obsolete comment.
7134 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7136         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7137         mpxchk, mpxld and mpxst types.
7138         (define_attr length_immediate): Remove all processing of mpx types.
7139         (define_attr prefix_0f): Ditto.
7140         (define_attr memory): Ditto.
7142 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7144         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7145         (log<mode>2): Change operand 1 predicate to general_operand.
7146         Extend operand 1 to XFmode and generate logxf3 insn.
7147         (log10<mode>2): Change operand 1 predicate to general_operand.
7148         Extend operand 1 to XFmode and generate log10xf3 insn.
7149         (log2<mode>2): Change operand 1 predicate to general_operand.
7150         Extend operand 1 to XFmode and generate log2xf3 insn.
7151         (fyl2xp1_extend<mode>xf3_i387): Remove.
7152         (log1p<mode>2): Change operand 1 predicate to general_operand.
7153         Extend operand 1 to XFmode and generate log1pxf3 insn.
7154         (fxtract_extend<mode>xf3_i387): Remove.
7155         (logb<mode>2): Change operand 1 predicate to general_operand.
7156         Extend operand 1 to XFmode and generate logbxf3 insn.
7157         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7158         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7159         (significand<mode>2): Change operand 1 predicate to general_operand.
7160         Extend operand 1 to XFmode and generate significandxf3 insn.
7162 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7164         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7165         (load_specs): Use fatal_error.
7166         (DELETE_IF_ORDINARY, process_command): Use error.
7167         (execute, run_attempt): Use fatal_error.
7169 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7171         * diagnostic-core.h (sorry_at): New prototype.
7172         * diagnostic.c (sorry_at): New function.
7174 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7176         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7177         by zero as VR_UNDEFINED.
7179 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7181         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7182         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7183         (<sincos>mode2): New expander.
7184         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7185         (sincos -> sin, cos splitters): Remove splitter patterns.
7186         (sincos<mode>3): Change operand 2 predicate to general_operand.
7187         Extend operand 2 to XFmode and generate sincosxf3 insn.
7188         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7189         Change operand 3 predicate to const1_operand.
7190         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7191         (tanxf2): Update operands in the call to fptanxf4_i387.
7192         (tan<mode>2): Change operand 1 predicate to general_operand.
7193         Extend operand 1 to XFmode and generate tanxf3 insn.
7194         (atan2xf3): Rename from *fpatanxf3_i387.
7195         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7196         (atan2xf3): Remove expander.
7197         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7198         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7199         (atan<mode>2): Change operand 1 predicate to general_operand.
7200         Extend operand 1 to XFmode and generate atanxf3 insn.
7202 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7204         * config/i386/i386.md (x87/SSE constant load splitter): Use
7205         memory_operand instead of nonimmediate_operand for input operand
7206         predicate.
7208 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7210         * config/i386/i386.md (float partial SSE register stall splitter): Move
7211         splitter near its instruction pattern.
7212         (float_extend partial SSE register stall splitter): Ditto.
7213         (float_truncate partial SSE register stall splitter): Ditto.
7215 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7217         PR target/86794
7218         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7219         to speculation_safe_value_not_needed.
7221         PR target/85666
7222         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7223         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7224         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7225         leaf_function_p, instead use has_hard_reg_initial_val.
7227 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7229         * gcc.h (pfatal_with_name): Don't declare here.
7230         * gcc.c (pfatal_with_name): Make static.
7232 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7234         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7235         earlyclobber.
7237 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7239         PR rtl-optimization/85458
7240         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7241         priority hook to reduce the priority of EXPR.
7243 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7245         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7246         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7247         with X87_ENABLE_FLOAT.  Remove preparation code.
7248         (*float<SWI48:mode><MODEF:mode>2): Rename from
7249         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7250         math using "enabled" attribute.
7251         (*floatdi<MODEF:mode>2_i387): Rename from
7252         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7253         enable for 32bit targets only.
7254         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7255         splitter.
7256         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7257         as operand 1 predicate.  Rewrite as define_insn_and_split.
7258         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7260 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7262         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7263         to fallthru to FLOAT case.
7265 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7267         PR target/86731
7268         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7269         around folding of vec_sl to handle out of range shift values.
7271 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7273         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7274         Update callers to gen_fix_trunc<mode>_i387_fisttp
7275         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7276         nonimmediate_operand.
7277         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7278         and corresponding splitters.
7279         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7280         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7281         (fix_truncdi_i387_with_temp): Remove insn pattern
7282         and corresponding splitters.
7283         (fix_trunc<mode>_i387): Change operand 0 predicate to
7284         nonimmediate_operand.
7285         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7286         and corresponding splitters.
7287         (*fistdi2_1): Remove.
7288         (fistdi2): Ditto.
7289         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7290         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7291         (*fist<mode>2_1): Remove.
7292         (fist<mode>2): Ditto.
7293         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7294         splitters.
7295         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7296         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7297         (fistdi2_<rounding>): Change operand 0 predicate to
7298         nonimmediate_operand.
7299         (fistdi2_<rounding>_with_temp): Remove insn pattern
7300         and corresponding splitters.
7301         (fist<mode>2_<rounding>): Change operand 0 predicate to
7302         nonimmediate_operand.
7303         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7304         and corresponding splitters.
7306         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7308 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7310         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7311         the init value.
7313 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7315         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7316         early gimple folding of vec_splat().
7317         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7318         * gimple-fold.h: Add an extern define for tree_vec_extract().
7320 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7322         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7323         wrappers around TREE_TYPE comparisons.
7325 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7327         PR target/80080
7328         * config/s390/predicates.md: Add nonsym_memory_operand.
7329         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7330         contains a SYMBOL_REF, load it into an intermediate pseudo.
7331         (s390_emit_compare_and_swap): Legitimize operand.
7332         * config/s390/s390.md: Use the new nonsym_memory_operand
7333         with UNSPECV_CAS patterns.
7335 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7337         PR target/80080
7338         * config/s390/s390-passes.def: New file.
7339         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7340         declaration.
7341         (make_pass_s390_early_mach): Add declaration.
7342         * config/s390/s390.c (make_pass_s390_early_mach):
7343         (s390_option_override): Remove dynamic registration.
7344         * config/s390/t-s390: Add s390-passes.def.
7346 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7348         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7349         Remove UNSPEC_LTREL_BASE check.
7350         (annotate_constant_pool_refs): Likewise.
7351         (find_constant_pool_ref): Likewise.
7352         (find_ltrel_base): Removed.
7353         (replace_ltrel_base): Removed.
7354         (s390_mainpool_finish): Remove replace_ltrel_base call.
7355         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7356         (s390_chunkify_finish): Remove replace_ltrel_base call.
7357         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7359 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7361         PR target/86779
7362         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7363         to speculation_safe_value_not_needed.
7365 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7366             Bernd Schmidt  <bernds_cb1@t-online.de>
7368         * config/nvptx/nvptx-opts.h: New file.
7369         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7370         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7371         (ASM_SPEC): Define.
7372         (TARGET_SM35): New macro.
7373         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7374         correct predicate.
7375         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7376         values.
7377         (misa=): New option.
7378         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7380 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7382         * config/i386/i386.md (truncdfsf2): Remove expander.
7383         (truncdfsf2_with_temp): Ditto.
7384         (truncxf<mode>2): Ditto.
7385         (*truncdfsf_fast_mixed): Remove insn pattern.
7386         (*truncdfsf_fast_i387): Ditto.
7387         (*truncdfsf_mixed): Ditto.
7388         (*truncdfsf_i387): Ditto.
7389         (*truncdfsf2_i387_1): Ditto.
7390         (*truncxfsf2_mixed): Ditto.
7391         (*truncxfdf2_mixed): Ditto.
7392         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7393         to call gen_truncxf<mode>2 instead.
7394         (*truncxf<mode>2_i387): Remove.
7395         (reg->reg splitters): Remove splitter pattern.
7396         (reg->mem splitters): Ditto.
7398         (truncdfsf2): New insn pattern.
7399         (truncxf<mode>2): Ditto.
7401 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7403         * tree-ssa-math-opts.c (is_mult_by): New function.
7404         (is_square_of): Use the above.
7405         (optimize_recip_sqrt): New function.
7406         (pass_cse_reciprocals::execute): Use the above.
7408 2018-09-05  Richard Biener  <rguenther@suse.de>
7410         PR bootstrap/87134
7411         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7412         to zero-init the emplaced vec.
7414 2018-09-05  Martin Liska  <mliska@suse.cz>
7416         PR tree-optimization/87205
7417         * tree-switch-conversion.c (pass_lower_switch::execute):
7418         Group cases for switch statements.
7420 2018-09-05  Richard Biener  <rguenther@suse.de>
7422         PR tree-optimization/87217
7423         * tree-ssa-sccvn.c (vuse_valueize): New.
7424         (vn_reference_lookup_pieces): Use it.
7425         (vn_reference_lookup): Likewise.
7427 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7429         PR c++/87137
7430         * stor-layout.c (place_field): Scan forwards to check last
7431         bitfield when ms_bitfield_placement is in effect.
7433 2018-09-05  Richard Biener  <rguenther@suse.de>
7435         PR bootstrap/87225
7436         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7437         return.
7439 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7440             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7442         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7443         * config.gcc (extra_objs): Build it.
7444         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7445         Likewise.
7446         * config/aarch64/aarch64-passes.def
7447         (pass_tag_collision_avoidance): New pass.
7448         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7449         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7450         (aarch64_classify_address): Remove static qualifier.
7451         (aarch64_address_info, aarch64_address_type): Move to...
7452         * config/aarch64/aarch64-protos.h: ... here.
7453         (make_pass_tag_collision_avoidance): New function.
7454         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7455         New tuning flag.
7457 2018-09-05  Martin Liska  <mliska@suse.cz>
7459         * doc/gcov.texi: Update documentation of humar
7460         readable mode.
7461         * gcov.c (format_count): Print one decimal place, it provides
7462         more fine number of situations like '1G' vs. '1.4G'.
7464 2018-09-05  Martin Liska  <mliska@suse.cz>
7466         PR target/87164
7467         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7468         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7469         to generate a MASK value.
7471 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7473         PR debug/86593
7474         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7475         if frame pointer isn't used.
7476         (compute_frame_pointer_to_fb_displacement): Likewise.
7478 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7480         PR target/87198
7481         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7482         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7483         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7484         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7485         and OPTION_MASK_ISA_XSAVEC_UNSET.
7487 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7489         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7490         XOR operations in NAND case.
7492 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7494         * wide-int-range.cc (wide_int_range_convert): New.
7495         * wide-int-range.h (wide_int_range_convert): New.
7496         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7497         code into wide_int_range_convert.
7498         (extract_range_into_wide_ints): Do not munge anti range constants
7499         into the entire domain.  Just return the range back.
7501 2018-09-04  Martin Liska  <mliska@suse.cz>
7503         * genmatch.c (output_line_directive): Add new argument
7504         fnargs.
7505         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7507 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7509         * doc/invoke.texi (Option Summary): Add whitespace.
7511         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7513 2018-09-04  Richard Biener  <rguenther@suse.de>
7515         PR tree-optimization/87211
7516         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7517         backedge value we're supposed to treat as VARYING also number
7518         the PHI to VARYING in case it got a different value-number already.
7520 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7522         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7523         (extract_range_from_binary_expr_1): Do not call
7524         vrp_can_optimize_bit_op.
7525         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7526         static.
7527         (wide_int_range_get_mask_and_bounds): New.
7528         (wide_int_range_optimize_bit_op): New.
7529         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7530         (wide_int_range_bit_and): Same.
7531         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7532         (wide_int_range_optimize_bit_op): New.
7533         (wide_int_range_get_mask_and_bounds): New.
7535 2018-09-04  Richard Biener  <rguenther@suse.de>
7537         PR tree-optimization/87176
7538         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7539         variable.  When value-numbering a virtual PHI node make sure
7540         to not value-number to the backedge value.
7542 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7544         * doc/extend.texi (Long Long, Hex Floats): Document support for
7545         long long and hex floats in more recent versions of ISO C++.
7547 2018-09-03  Richard Biener  <rguenther@suse.de>
7549         PR tree-optimization/87177
7550         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7551         cleanup.
7553 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7555         * bb-reorder.c (edge_order): Convert to C-qsort-style
7556         tri-state comparator.
7557         (reorder_basic_blocks_simple): Change std::stable_sort to
7558         gcc_stablesort.
7560 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7562         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7563         tri-state comparator.
7564         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7566 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7568         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7569         (mergesort): ... here as maximum count for using netsort.
7570         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7571         (gcc_stablesort): New.
7572         * system.h (gcc_stablesort): Declare.
7574 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7576         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7577         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7578         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7580 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7582         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7583         lxsdx and stxsdx alternatives.
7584         (*mov<mode>_hardfloat64): Ditto.
7585         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7587 2018-09-03  Richard Biener  <rguenther@suse.de>
7589         PR tree-optimization/87200
7590         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7591         simplify result.
7593 2018-09-03  Martin Liska  <mliska@suse.cz>
7595         PR tree-optimization/87201
7596         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7597         Fix parenthesis in an expression.
7599 2018-09-03  Richard Biener  <rguenther@suse.de>
7601         PR tree-optimization/87197
7602         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7603         visited.  CSE the VN_INFO hashtable lookup.
7605         PR tree-optimization/87169
7606         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7607         iterating make sure there's no extra backedges from irreducible
7608         regions feeding the header.  Mark the destination block
7609         executable.
7611 2018-09-03  Martin Liska  <mliska@suse.cz>
7613         PR driver/83193
7614         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7615         * common/common-targhooks.c (default_get_valid_option_values):
7616         New function.
7617         * common/common-targhooks.h (default_get_valid_option_values):
7618         Likewise.
7619         * common/config/i386/i386-common.c: Move processor_target_table
7620         from i386.c.
7621         (ix86_get_valid_option_values): New function.
7622         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7623         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7624         (PTA_*): Move all defined masks into i386-common.c.
7625         (ix86_function_specific_restore): Use new processor_cost_table.
7626         * config/i386/i386.h (struct ptt): Moved from i386.c.
7627         (struct pta): Likewise.
7628         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7629         * doc/tm.texi.in: Likewise.
7630         * opt-suggestions.c (option_proposer::suggest_option):
7631         Pass prefix to build_option_suggestions.
7632         (option_proposer::get_completions): Likewise.
7633         (option_proposer::build_option_suggestions): Use the new target
7634         hook.
7635         * opts.c (struct option_help_tuple): New struct.
7636         (print_filtered_help): Use the new target hook.
7638 2018-09-03  Martin Liska  <mliska@suse.cz>
7640         PR middle-end/59521
7641         * predict.c (set_even_probabilities): Add likely_edges
7642         argument and handle cases where we have precisely one
7643         likely edge.
7644         (combine_predictions_for_bb): Catch also likely_edges.
7645         (tree_predict_by_opcode): Handle gswitch statements.
7646         * tree-cfg.h (find_case_label_for_value): New declaration.
7647         (find_taken_edge_switch_expr): Likewise.
7648         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7649         Find pivot in decision tree based on probabily, not by number of
7650         nodes.
7652 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7654         * doc/standards.texi (Standards): Update Objective-C reference.
7656 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7658         * doc/install.texi (Prerequisites): Update link for MPC.
7660 2018-09-01  Michael Matz  <matz@suse.de>
7662         PR tree-optimization/87074
7663         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7664         PHIs for outer-loop uses.
7666 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7668         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7669         * doc/invoke.texi (C Dialect Options): Ditto.
7671 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7673         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7675 2018-08-31  Richard Biener  <rguenther@suse.de>
7677         PR tree-optimization/87168
7678         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7679         (rpo_elim::eliminate_avail): When OP was not visited it must
7680         be available.
7682 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7684         * tree-vrp.c (copy_value_range): Convert param "from" from
7685         "value_range *" to "const value_range *".
7686         (range_is_null): Likewise for param "vr".
7687         (range_int_cst_p): Likewise.
7688         (range_int_cst_singleton_p): Likewise.
7689         (symbolic_range_p): Likewise.
7690         (value_ranges_intersect_p): Likewise for both params.
7691         (value_range_nonnegative_p): Likewise for param "vr".
7692         (value_range_constant_singleton): Likewise.
7693         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7694         (extract_range_into_wide_ints): Likewise for param "vr".
7695         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7696         and "vr1".
7697         (vrp_can_optimize_bit_op): Likewise.
7698         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7699         "vr1_".
7700         (extract_range_from_unary_expr): Likewise.
7701         (debug_value_range): Likewise for param "vr".
7702         (value_range::dump): Add "const" qualifier.
7703         (vrp_prop::check_array_ref): Convert local "vr" from
7704         "value_range *" to "const value_range *".
7705         (vrp_prop::check_mem_ref): Likewise.
7706         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7707         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7708         (vrp_intersect_ranges): Likewise.
7709         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7710         (vrp_prop::vrp_finalize): Likewise.
7711         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7712         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7713         (extract_range_from_unary_expr): Likewise.
7714         (value_range_constant_singleton): Likewise.
7715         (symbolic_range_p): Likewise.
7716         (copy_value_range): Likewise.
7717         (extract_range_from_binary_expr_1): Likewise.
7718         (range_int_cst_p): Likewise.
7719         (vrp_set_zero_nonzero_bits): Likewise.
7720         (range_int_cst_singleton_p): Likewise.
7722 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7724         * config/aarch64/arm_neon.h (vabsd_s64): New.
7725         (vnegd_s64): Likewise.
7727 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7729         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7731 2018-08-31  Martin Liska  <mliska@suse.cz>
7733         * ipa-icf.c (sem_item::add_type): Use
7734         sem_item::m_type_hash_cache.
7735         * ipa-icf.h: Move the cache from sem_item_optimizer
7736         to sem_item.
7738 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7740         * doc/extend.texi (Backwards Compatibility): Remove implicit
7741         extern C leeway of () being (...).
7743 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7745         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7747 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7749         PR target/86684
7750         PR target/87149
7751         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7753 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7755         PR middle-end/87138
7756         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7757         gen_int_mode.  Formatting fixes.
7759 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7761         * target.def (custom_function_descriptors): Improve documentation.
7762         * doc/tm.texi.in (Trampolines): Expand discussion of function
7763         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7764         beginning of the section.
7765         * doc/tm.texi: Regenerated.
7767 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7769         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7770         base class in the initializer list.  This is a workaround for
7771         building with older GCC.
7772         (class auto_bb_flag): Likewise.
7774 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7776         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7777         (altivec_vcmpequ<VI_char>_p): Remove star.
7778         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7779         vector load modes.
7780         (expand_strncmp_vec_sequence): New function.
7781         (emit_final_str_compare_vec): New function.
7782         (expand_strn_compare): Add support for vector strncmp.
7783         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7784         length specification to bytes.
7785         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7786         (vcmpnezb_p): New pattern.
7787         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7788         for option -mstring-compare-inline-limit.
7790 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7792         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7793         (PTA_SKYLAKE): Add PTA_AES.
7794         (PTA_GOLDMONT): Likewise.
7796 2018-08-29  Jan Hubicka  <jh@suse.cz>
7798         PR lto/86517
7799         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7800         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7802 2018-08-29  Jan Hubicka  <jh@suse.cz>
7804         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7805         TYPE_STUB_DECL.
7806         (hash_tree): Do not visit TYPE_STUB_DECL.
7807         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7808         stream TYPE_STUB_DECL.
7809         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7810         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7811         after free_lang_data.
7812         (type_in_anonymous_namespace_p): Likewise.
7814 2018-08-29  Jan Hubicka  <jh@suse.cz>
7816         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7817         comment that it has to be even number.
7818         (class sreal): Change m_sig type to int32_t.
7819         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7820         int64_t for temporary calculations.
7821         (sreal_verify_basics): Drop one bit from minimum and maximum.
7823 2018-08-30  Richard Biener  <rguenther@suse.de>
7825         PR tree-optimization/87147
7826         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7827         (visit_phi): When the degenerate result is from the backedge and
7828         we didn't visit its definition yet drop to VARYING.
7829         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7831 2018-08-29  Jan Hubicka  <jh@suse.cz>
7833         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7834         DECL_VINDEX.
7835         (hash_tree): Likewise.
7837 2018-08-29  Jan Hubicka  <jh@suse.cz>
7839         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7840         and TYPE_NEXT_REF_TO.
7842 2018-08-29  Jan Hubicka  <jh@suse.cz>
7844         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7845         comment that it has to be even number.
7846         (class sreal): Change m_sig type to int32_t.
7847         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7848         int64_t for temporary calculations.
7849         (sreal_verify_basics): Drop one bit from minimum and maximum.
7851 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7853         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7855 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7857         PR middle-end/86995
7858         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7859         if to_add is negative.
7861 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7863         PR middle-end/87053
7864         * builtins.c (c_strlen): Improve range checks.
7866 2018-08-29  Martin Sebor  <msebor@redhat.com>
7867             Jeff Law  <law@redhat.com>
7869         PR tree-optimization/86714
7870         PR tree-optimization/86711
7871         * builtins.c (c_strlen): Add arguments to call to string_constant.
7872         * expr.c (string_constant): Add argument.  Detect missing nul
7873         terminator and outermost declaration it's missing in.
7874         * expr.h (string_constant): Add argument.
7875         * fold-const.c (read_from_constant_string): Add arguments to call to
7876         string_constant.
7877         (c_getstr): Likewise.
7878         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7879         to string_constant.
7880         * tree-ssa-strlen.c (get_stridx): Likewise.
7882 2018-08-29  Jan Hubicka  <jh@suse.cz>
7884         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7885         Do not stream DECL_VINDEX.
7886         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7887         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7888         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7890 2018-08-29  Richard Biener  <rguenther@suse.de>
7892         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7893         virtual operands that are not default defs to honor region
7894         boundaries.
7895         (rpo_vn_valueize): Remove ineffective code here.
7897 2018-08-29  Richard Biener  <rguenther@suse.de>
7899         PR tree-optimization/87132
7900         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7901         when skipping defs reachable over backedges.
7903 2018-08-29  Richard Biener  <rguenther@suse.de>
7905         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7906         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7907         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7908         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7909         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7910         (vn_reference_lookup_3): Remove use of const_parms.
7911         (free_rpo_vn): Do not free const_parms.
7912         (do_rpo_vn): Do not call init_const_parms.
7913         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7914         SSA_NAME_POINTS_TO_READONLY_MEMORY.
7915         (call_may_clobber_ref_p_1): Likewise.
7917 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
7919         PR other/86726
7920         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7921         (-O): Ditto.
7922         (-ftree-scev-cprop): Document.
7924 2018-08-29  Jan Hubicka  <jh@suse.cz>
7926         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7927         parameters.
7928         (sreal constructor): Update.
7929         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7930         sreal:operator/): Update.
7932 2018-08-29  Martin Liska  <mliska@suse.cz>
7934         * tree-switch-conversion.c (switch_conversion::expand):
7935         Strenghten assumption about gswitch statements.
7937 2018-08-29  Richard Biener  <rguenther@suse.de>
7939         PR tree-optimization/87117
7940         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7941         re-value-number released SSA VDEFs.
7943 2018-08-29  Richard Biener  <rguenther@suse.de>
7945         PR tree-optimization/87126
7946         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7948 2018-08-28  Jim Wilson  <jimw@sifive.com>
7950         * config/riscv/pic.md: Rewrite.
7951         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7952         invalid address.
7953         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7954         (SOFTF, default_load, softload, softstore): New.
7956 2018-08-28  Jeff Law  <law@redhat.com>
7958         * fold-const.c (fold_binary_loc): Remove recently added assert.
7960 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
7962         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7963         to OP parmeter of generated function.
7965 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
7967         PR tree-optimization/87009
7968         * match.pd: Add boolean optimizations.
7970 2018-08-28  Martin Sebor  <msebor@redhat.com>
7972         PR middle-end/86631
7973         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7974         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7975         (pass_walloca::gate): Use it.
7976         (alloca_call_type): Same.
7977         (pass_walloca::execute): Same.
7978         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7980 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
7982         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7983         GCC_VERSION for usage of "__gcc_dump_printf__" format from
7984         >= 3005 to >= 9000.
7986 2018-08-28  Richard Biener  <rguenther@suse.de>
7988         PR tree-optimization/87124
7989         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7990         constants before looking up avail.
7992 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
7994         PR middle-end/87099
7995         * calls.c (maybe_warn_nonstring_arg): Punt early if
7996         warn_stringop_overflow is zero.  Don't call get_range_strlen
7997         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7998         Swap comparison operands to have constants on rhs.  Only use
7999         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8000         increment lenrng[0].
8002 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8004         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8005         use of tree_to_shwi.  Remove duplicated test for the size being
8006         a whole number of bytes.
8008 2018-08-28  Richard Biener  <rguenther@suse.de>
8010         PR tree-optimization/87117
8011         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8012         Handle removed stmt without LHS (GIMPLE_NOP).
8014 2018-08-28  Richard Biener  <rguenther@suse.de>
8016         PR tree-optimization/87117
8017         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8018         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8020 2018-08-28  Richard Biener  <rguenther@suse.de>
8022         PR tree-optimization/87117
8023         * tree-ssa-pre.c (compute_avail): Do not make expressions
8024         with predicated values available.
8025         (get_expr_value_id): Assert we do not run into predicated value
8026         expressions.
8028 2018-08-28  Richard Biener  <rguenther@suse.de>
8030         PR tree-optimization/87117
8031         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8032         get virtual operands.
8033         (get_expr_operands): Handle STRING_CST like other decls.
8035 2018-08-28  Martin Liska  <mliska@suse.cz>
8037         * tree.h: Update documentation of fndecl_built_in_p
8038         functions.
8041 2018-08-27  Jeff Law  <law@redhat.com>
8042         PR tree-optimization/87110
8043         * tree-ssa-dse.c (compute_trims): Handle non-constant
8044         TYPE_SIZE_UNIT.
8046 2018-08-27  Martin Sebor  <msebor@redhat.com>
8048         PR tree-optimization/86914
8049         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8051 2018-08-27  Martin Sebor  <msebor@redhat.com>
8053         PR tree-optimization/87112
8054         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8055         the type of the bound argument.
8057 2018-08-27  Jeff Law  <law@redhat.com>
8059         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8060         type does not have a TYPE_SIZE_UNIT.
8062 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8064         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8065         with include of backend.h.
8067 2018-08-27  Richard Biener  <rguenther@suse.de>
8069         PR tree-optimization/86927
8070         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8071         use const cond reduction code.
8073 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8075         PR tree-optimization/85758
8076         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8078 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8080         PR c++/87091
8081         * diagnostic-show-locus.c (class layout_range): Update for
8082         conversion of show_caret_p to a tri-state.
8083         (layout_range::layout_range): Likewise.
8084         (make_range): Likewise.
8085         (layout::maybe_add_location_range): Likewise.
8086         (layout::should_print_annotation_line_p): Don't show annotation
8087         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8088         (layout::get_state_at_point): Update for conversion of
8089         show_caret_p to a tri-state.  Bail out early for
8090         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8091         underlining or source colorization.
8092         (gcc_rich_location::add_location_if_nearby): Update for conversion
8093         of show_caret_p to a tri-state.
8094         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8095         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8096         Likewise.
8097         (selftest::test_one_liner_labels): Likewise.
8098         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8099         conversion of show_caret_p to a tri-state.
8100         * pretty-print.c (text_info::set_location): Likewise.
8101         * pretty-print.h (text_info::set_location): Likewise.
8102         * substring-locations.c (format_warning_n_va): Likewise.
8103         * tree-diagnostic.c (default_tree_printer): Likewise.
8104         * tree-pretty-print.c (newline_and_indent): Likewise.
8106 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8108         PR c++/87091
8109         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8110         line above for line-insertion fix-it hints.
8111         (selftest::test_fixit_insert_containing_newline): Update the
8112         expected results, and add a test with line-numbering enabled.
8114 2018-08-27  Martin Liska  <mliska@suse.cz>
8116         PR sanitizer/86962
8117         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8118         params with DECL_HAS_VALUE_EXPR_P.
8120 2018-08-27  Martin Liska  <mliska@suse.cz>
8122         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8123         selected expansion strategy.
8125 2018-08-27  Martin Liska  <mliska@suse.cz>
8127         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8128         * builtins.c (is_builtin_fn): Likewise.
8129         * attribs.c (diag_attr_exclusions): Use new function
8130         fndecl_built_in_p and remove check for FUNCTION_DECL if
8131         possible.
8132         (builtin_mathfn_code): Likewise.
8133         (fold_builtin_expect): Likewise.
8134         (fold_call_expr): Likewise.
8135         (fold_builtin_call_array): Likewise.
8136         (fold_call_stmt): Likewise.
8137         (set_builtin_user_assembler_name): Likewise.
8138         (is_simple_builtin): Likewise.
8139         * calls.c (gimple_alloca_call_p): Likewise.
8140         (maybe_warn_nonstring_arg): Likewise.
8141         * cfgexpand.c (expand_call_stmt): Likewise.
8142         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8143         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8144         (cgraph_node::verify_node): Likewise.
8145         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8146         (cgraph_node::create_clone): Likewise.
8147         * config/arm/arm.c (arm_insert_attributes): Likewise.
8148         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8149         * dse.c (scan_insn): Likewise.
8150         * expr.c (expand_expr_real_1): Likewise.
8151         * fold-const.c (operand_equal_p): Likewise.
8152         (fold_binary_loc): Likewise.
8153         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8154         * gimple-low.c (lower_stmt): Likewise.
8155         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8156         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8157         Likewise.
8158         * gimple.c (gimple_build_call_from_tree): Likewise.
8159         (gimple_call_builtin_p): Likewise.
8160         (gimple_call_combined_fn): Likewise.
8161         * gimplify.c (gimplify_call_expr): Likewise.
8162         (gimple_boolify): Likewise.
8163         (gimplify_modify_expr): Likewise.
8164         (gimplify_addr_expr): Likewise.
8165         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8166         * ipa-cp.c (determine_versionability): Likewise.
8167         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8168         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8169         * ipa-split.c (visit_bb): Likewise.
8170         (split_function): Likewise.
8171         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8172         * lto-cgraph.c (input_node): Likewise.
8173         * lto-streamer-out.c (write_symbol): Likewise.
8174         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8175         (lower_omp_1): Likewise.
8176         * predict.c (strip_predict_hints): Likewise.
8177         * print-tree.c (print_node): Likewise.
8178         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8179         * trans-mem.c (is_tm_irrevocable): Likewise.
8180         (is_tm_load): Likewise.
8181         (is_tm_simple_load): Likewise.
8182         (is_tm_store): Likewise.
8183         (is_tm_simple_store): Likewise.
8184         (is_tm_abort): Likewise.
8185         (tm_region_init_1): Likewise.
8186         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8187         * tree-cfg.c (verify_gimple_call): Likewise.
8188         (move_stmt_r): Likewise.
8189         (stmt_can_terminate_bb_p): Likewise.
8190         * tree-eh.c (lower_eh_constructs_2): Likewise.
8191         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8192         * tree-inline.c (remap_gimple_stmt): Likewise.
8193         (copy_bb): Likewise.
8194         (estimate_num_insns): Likewise.
8195         (fold_marked_statements): Likewise.
8196         * tree-sra.c (scan_function): Likewise.
8197         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8198         (optimize_stack_restore): Likewise.
8199         (pass_fold_builtins::execute): Likewise.
8200         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8201         (mark_all_reaching_defs_necessary_1): Likewise.
8202         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8203         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8204         (pass_forwprop::execute): Likewise.
8205         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8206         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8207         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8208         * tree-ssa-strlen.c (get_string_length): Likewise.
8209         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8210         (find_func_aliases_for_call): Likewise.
8211         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8212         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8213         * tree-tailcall.c (find_tail_calls): Likewise.
8214         * tree.c (need_assembler_name_p): Likewise.
8215         (free_lang_data_in_decl): Likewise.
8216         (get_call_combined_fn): Likewise.
8217         * ubsan.c (is_ubsan_builtin_p): Likewise.
8218         * varasm.c (incorporeal_function_p): Likewise.
8219         * tree.h (DECL_BUILT_IN): Remove and replace with
8220         fndecl_built_in_p.
8221         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8222         (fndecl_built_in_p): New.
8224 2018-08-27  Martin Liska  <mliska@suse.cz>
8226         PR tree-optimization/86847
8227         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8228         Dump also subtree probability.
8229         (switch_decision_tree::do_jump_if_equal): New function.
8230         (switch_decision_tree::emit_case_nodes): Handle special
8231         situations in balanced tree that can be emitted much simpler.
8232         Fix calculation of probabilities that happen in tree expansion.
8233         * tree-switch-conversion.h (struct cluster): Add
8234         is_single_value_p.
8235         (struct simple_cluster): Likewise.
8236         (struct case_tree_node): Add new function has_child.
8237         (do_jump_if_equal): New.
8239 2018-08-27  Martin Liska  <mliska@suse.cz>
8241         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8242         Add new argument to bit_test_cluster constructor.
8243         (bit_test_cluster::emit): Set bits really number of values
8244         handlel by a test.
8245         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8246         probability argument.
8247         * tree-switch-conversion.h (struct bit_test_cluster):
8248         Add m_handles_entire_switch.
8250 2018-08-27  Martin Liska  <mliska@suse.cz>
8252         PR tree-optimization/86702
8253         * tree-switch-conversion.c (jump_table_cluster::emit):
8254         Make probabilities even for values in jump table
8255         according to number of cases handled.
8256         (switch_decision_tree::compute_cases_per_edge): Pass
8257         argument to reset_out_edges_aux function.
8258         (switch_decision_tree::analyze_switch_statement): Likewise.
8259         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8260         Make it static.
8262 2018-08-27  Martin Liska  <mliska@suse.cz>
8264         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8265         cfun argument explicitly.
8266         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8267         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8268         function gimple_switch_default_bb.
8269         (convert_switch_statements):
8270         (expand_builtins):
8271         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8272         * stmt.c (label_to_block_fn): Use label_to_block and pass
8273         cfun argument explicitly and use gimple_switch_label_bb.
8274         (expand_case): Likewise.
8275         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8276         cfun argument explicitly. Likewise.
8277         (make_edges_bb): Likewise.
8278         (make_cond_expr_edges): Likewise.
8279         (get_cases_for_edge): Likewise.
8280         (make_gimple_switch_edges): Likewise.
8281         (label_to_block_fn): Likewise.
8282         (label_to_block): Likewise.
8283         (make_goto_expr_edges): Likewise.
8284         (make_gimple_asm_edges): Likewise.
8285         (main_block_label): Likewise.
8286         (group_case_labels_stmt): Likewise.
8287         (find_taken_edge_computed_goto): Likewise.
8288         (find_taken_edge_switch_expr): Likewise.
8289         (gimple_verify_flow_info): Likewise.
8290         (gimple_redirect_edge_and_branch): Likewise.
8291         (gimple_switch_label_bb): New function.
8292         (gimple_switch_default_bb): Likewise.
8293         (gimple_switch_edge): Likewise.
8294         (gimple_switch_default_edge): Likewise.
8295         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8296         (label_to_block): ... with this.
8297         (gimple_switch_label_bb): New.
8298         (gimple_switch_default_bb): Likewise.
8299         (gimple_switch_edge): Likewise.
8300         (gimple_switch_default_edge): Likewise.
8301         * tree-cfgcleanup.c (convert_single_case_switch): Use
8302         new gimple functions and pass new argument to label_to_block.
8303         (cleanup_control_flow_bb):
8304         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8305         cfun argument explicitly.
8306         (make_eh_edges): Likewise.
8307         (redirect_eh_dispatch_edge): Likewise.
8308         (lower_resx): Likewise.
8309         (lower_eh_dispatch): Likewise.
8310         (maybe_remove_unreachable_handlers): Likewise.
8311         (unsplit_eh): Likewise.
8312         (cleanup_empty_eh): Likewise.
8313         (verify_eh_edges): Likewise.
8314         (verify_eh_dispatch_edge): Likewise.
8315         * tree-ssa-dom.c (record_edge_info): Likewise.
8316         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8317         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8318         (thread_through_normal_block): Likewise.
8319         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8320         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8321         * tree-switch-conversion.c (switch_conversion::collect): Use new
8322         gimple functions.
8323         (switch_conversion::check_final_bb): Likewise.
8324         (switch_conversion::gather_default_values): Pass new argument
8325         to label_to_block.
8326         (switch_conversion::build_constructors): Likewise.
8327         (switch_decision_tree::compute_cases_per_edge): Use new
8328         gimple_switch_edge function.
8329         (switch_decision_tree::analyze_switch_statement): Pass new argument
8330         to label_to_block.
8331         (switch_decision_tree::try_switch_expansion): Use
8332         gimple_switch_default_edge.
8333         * tree-vrp.c (find_switch_asserts): Pass new argument
8334         to label_to_block.
8335         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8336         (vr_values::simplify_switch_using_ranges): Likewise.
8338 2018-08-27  Richard Biener  <rguenther@suse.de>
8340         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8341         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8343         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8344         (struct vn_nary_op_s): Add unwind_to member.  Add
8345         predicated_values flag and put result into a union together
8346         with a linked list of vn_pval.
8347         (struct vn_ssa_aux): Add name member to make maintaining
8348         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8349         needed info, dfsnum, low, visited, on_sccstack, use_processed
8350         and range_info_anti_range_p members.
8351         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8352         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8353         New functions.
8354         (vn_valueize): New global.
8355         (vn_context_bb): Likewise.
8356         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8357         VN_INFO_PTR_INFO): Remove.
8358         * tree-ssa-sccvn.c: ... (rewrite)
8359         (pass_fre::execute): For -O2+ initialize loops and run
8360         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8361         run RPO VN in non-optimistic mode.
8362         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8363         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8364         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8365         (rpo-vn-max-loop-depth): Document.
8366         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8367         when valuezing the VUSE signals we walked out of the region.
8368         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8369         (phi_translate): Set VN context block to use for availability
8370         lookup.
8371         (compute_avail): Likewise.
8372         (pre_valueize): New function.
8373         (pass_pre::execute): Adjust to the RPO VN API.
8375         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8376         (propagate_constants_for_unrolling): Remove.
8377         (tree_unroll_loops_completely): Perform value-numbering
8378         on the unrolled bodies loop parent.
8380 2018-08-27  Richard Biener  <rguenther@suse.de>
8382         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8383         for partial antic compute.
8385 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8387         PR rtl-optimization/87065
8388         * combine.c (simplify_if_then_else): Formatting fix.
8389         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8390         check.
8391         (known_cond): Don't return const_true_rtx for vector modes.  Use
8392         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8394 2018-08-27  Martin Liska  <mliska@suse.cz>
8396         PR gcov-profile/87069
8397         * gcov.c (process_file): Record files already processed
8398         and warn about a file being processed multiple times.
8400 2018-08-27  Martin Liska  <mliska@suse.cz>
8402         PR driver/83193
8403         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8404         Set default values for x_aarch64_*_string strings.
8405         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8406         prefix.  For -mabi do not print '=ABI' in help and use
8407         <option_value> format for -msve-vector-bits and -moverride
8408         options.
8410 2018-08-26  Jeff Law  <law@redhat.com>
8412         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8413         "cfg.h"
8415 2018-08-26  Marek Polacek  <polacek@redhat.com>
8417         PR c++/87029, Implement -Wredundant-move.
8418         * doc/invoke.texi: Document -Wredundant-move.
8420 2018-08-25  Martin Sebor  <msebor@redhat.com>
8422         PR tree-optimization/87059
8423         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8424         to the same type as the other.
8425         * fold-const.c (fold_binary_loc): Assert expectation.
8427 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8429         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8430         extraneous parentheses, dead code section and formatting.
8432 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8434         PR c++/87091
8435         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8436         wide enough for jumps in the line-numbering to be visible.
8437         (layout::print_gap_in_line_numbering): New member function.
8438         (layout::calculate_line_spans): When using line numbering, merge
8439         line spans that are only 1 line apart.
8440         (diagnostic_show_locus): When printing line numbers, show gaps in
8441         line numbering directly, rather than printing headers.
8442         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8443         line-numbering with multiple line spans.
8444         (selftest::test_fixit_insert_containing_newline_2): Add test of
8445         line-numbering, in which the spans are close enough to be merged.
8447 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8449         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8450         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8451         * tree-vrp.c (range_is_nonnull): Remove.
8452         (range_includes_zero_p): Accept value_range instead of min/max.
8453         (extract_range_from_binary_expr_1): Do not early bail on
8454         POINTER_PLUS_EXPR.
8455         Use range_includes_zero_p instead of range_is_nonnull.
8456         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8457         of range_is_nonnull.
8458         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8459         special case VR_ANTI_RANGE.
8460         (vrp_finalize): Same.
8461         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8462         instead of min/max.
8463         (range_is_nonnull): Remove.
8464         * vr-values.c (vrp_stmt_computes_nonzero): Use
8465         range_includes_zero_p instead of range_is_nonnull.
8466         (extract_range_basic): Pass value_range to range_includes_zero_p
8467         instead of range_is_nonnull.
8469 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8471         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8472         * emit-rtl.h (rtl_data): Remove return_bnd.
8473         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8474         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8475         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8476         (POINTER_BOUNDS_MODE): Remove definition.
8477         (make_pointer_bounds_mode): Remove.
8478         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8479         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8480         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8481         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8482         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8483         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8484         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8486         * config/i386/i386-modes.def (BND32, BND64): Remove.
8487         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8488         (dbx64_register_map): Ditto.
8489         (svr4_dbx_register_map): Ditto.
8490         (indirect_thunk_bnd_needed): Remove.
8491         (indirect_thunks_bnd_used): Ditto.
8492         (indirect_return_bnd_needed): Ditto.
8493         (indirect_return_via_cx_bnd): Ditto.
8494         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8495         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8496         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8497         (output_indirect_thunk_function): Remove handling of
8498         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8499         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8500         (ix86_save_reg): Remove handling of crtl->return_bnd.
8501         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8502         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8503         and UNSPEC_BNDLX_ADDR.
8504         (ix86_output_indirect_branch_via_reg): Remove handling of
8505         indirect_thunk_prefix_bnd.
8506         (ix86_output_indirect_branch_via_push): Ditto.
8507         (ix86_output_function_return): Ditto.
8508         (ix86_output_indirect_function_return): Ditto.
8509         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8510         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8511         (CALL_USED_REGISTERS): Ditto.
8512         (REG_ALLOC_ORDER): Update for removal of bound registers.
8513         (HI_REGISTER_NAMES): Ditto.
8514         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8515         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8516         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8517         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8518         (FIRST_PSEUDO_REG): Update.
8519         (BND): Remove mode iterator.
8520         * config/i386/predicates.md (bnd_mem_operator): Remove.
8522 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8524         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8525         vectors.
8527 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8529         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8530         the case in which the permute needs only a single element and
8531         repeats for every vector of the result.  Extend that case to
8532         handle variable-length vectors.
8533         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8535 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8537         PR debug/79342
8538         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8539         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8541 2018-08-24  Richard Biener  <rguenther@suse.de>
8543         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8544         bb_flags_allocated members.
8545         (auto_flag): New RAII class for allocating flags.
8546         (auto_edge_flag): New RAII class for allocating edge flags.
8547         (auto_bb_flag): New RAII class for allocating bb flags.
8548         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8549         flag dynamically.
8550         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8551         in favor of temporarily allocated BB flag.
8552         * hsa-brig.c: Re-order includes.
8553         * hsa-dump.c: Likewise.
8554         * hsa-regalloc.c: Likewise.
8555         * print-rtl.c: Likewise.
8556         * profile-count.c: Likewise.
8558 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8560         PR target/86989
8561         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8562         the TOC register.
8564 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8566         PR 87073/bootstrap
8567         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8568         from wide_int_range_multiplicative_op.
8570 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8572         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8573         "permutaion".
8575 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8577         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8578         to 'expanded'.
8580 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8582         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8583         full GENERIC expression used for replacement.
8585 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8587         * tree-vrp.c (abs_extent_range): Remove.
8588         (extract_range_into_wide_ints): Pass wide ints by reference.
8589         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8590         Pass wide ints by reference in all calls to
8591         extract_range_into_wide_ints.
8592         * wide-int-range.cc (wide_int_range_div): New.
8593         * wide-int-range.h (wide_int_range_div): New.
8594         (wide_int_range_includes_zero_p): New.
8595         (wide_int_range_zero_p): New.
8597 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8599         * config/aarch64/aarch64.md (arches): New enum.
8600         (arch): New enum attr.
8601         (arch_enabled): New attr.
8602         (enabled): Now uses arch_enabled only.
8603         (simd, sve, fp16): Removed attribute.
8604         (fp): Attr now defined in terms of 'arch'.
8605         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8606         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8607         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8608         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8609         attributes into 'arch'.
8610         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8611         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8612         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8613         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8614         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8615         'simd' attribute into 'arch'.
8616         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8617         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8618         Convert use of 'fp' attribute to 'arch'.
8619         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8620         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8621         into 'arch'.
8622         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8623         (different modes) Merge 'fp' and 'simd' into 'arch'.
8624         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8625         'simd' into 'arch'.
8627 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8629         PR rtl-optimization/87026
8630         * expmed.c (canonicalize_comparison): If we can no longer create
8631         pseudoregisters, don't.
8633 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8635         PR target/86951
8636         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8637         prototype.
8638         * config/arm/arm.c (speculation_barrier_libfunc): New static
8639         variable.
8640         (arm_init_libfuncs): Initialize it.
8641         (arm_emit_speculation_barrier): New function.
8642         * config/arm/arm.md (speculation_barrier): Call
8643         arm_emit_speculation_barrier for architectures that do not have
8644         DSB or ISB.
8645         (speculation_barrier_insn): Only match on Armv7 or later.
8647 2018-08-23  Richard Biener  <rguenther@suse.de>
8649         PR middle-end/87024
8650         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8651         calls.
8653 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8655         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8656         of single-vector TBLs.
8657         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8658         one input is given.
8660 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8662         PR target/85910
8663         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8664         aarch64_evpc_tbl guard.
8666 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8668         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8669         behaviour.
8671 2018-08-22  Martin Sebor  <msebor@redhat.com>
8673         PR middle-end/87052
8674         * tree-pretty-print.c (pretty_print_string): Add argument.
8675         (dump_generic_node): Call to pretty_print_string with string size.
8677 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8679         PR rtl-optimization/86771
8680         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8681         of two SETs into those two SETs, one to be placed at i2, if that SETs
8682         destination is modified between i2 and i3.
8684 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8686         PR tree-optimization/86725
8687         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8688         function.
8689         (vect_analyze_scalar_cycles_1): Check it.
8691 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8693         PR tree-optimization/86725
8694         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8695         an outer loop phi as a double reduction, make sure that the
8696         single user of the phi result is an inner loop phi.
8698 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8700         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8701         grouped stores with gaps to a strided group.
8703 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8705         * tree-vect-stmts.c (get_group_load_store_type)
8706         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8707         first statement in a group.
8709 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8711         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8712         the sequence used in gcc/gcc.c.
8714 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8716         PR other/704
8717         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8718         building it.
8720 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8722         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8723         Darwin10-specific unwinder-shim.
8724         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8725         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8726         New to cater for Darwin10 Rosetta.
8728 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8730         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8731         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8733 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8735         PR bootstrap/81033
8736         PR target/81733
8737         PR target/52795
8738         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8739         (dwarf2out_switch_text_section): Generate a local label for the
8740         second function sub-section and apply it as the second FDE start
8741         label.
8742         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8743         second sub-section start.
8745 2018-08-22  Richard Biener  <rguenther@suse.de>
8747         PR tree-optimization/86988
8748         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8750 2018-08-22  Richard Biener  <rguenther@suse.de>
8752         PR tree-optimization/86945
8753         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8755 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8757         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8758         a comment about how uses of r2 for .sdata2 come about.
8760 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8762         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8764 2018-08-21  Marek Polacek  <polacek@redhat.com>
8766         PR c++/86981, Implement -Wpessimizing-move.
8767         * doc/invoke.texi: Document -Wpessimizing-move.
8769 2018-08-21  Jan Hubicka  <jh@suse.cz>
8771         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8772         * tree.h (is_redundant_typedef): Remove.
8773         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8775 2018-08-21  Jan Hubicka  <jh@suse.cz>
8777         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8778         when possible.
8780 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8782         * expmed.c (extract_low_bits): Reject invalid subregs early.
8784 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8786         PR middle-end/86121
8787         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8788         behaviour.
8790 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8792         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8793         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8794         * config/vxworks.c: Likewise.
8796 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8798         * config/vxworks.c: Set targetm.have_ctors_dtors
8799         if HAVE_INITFINI_ARRAY_SUPPORT.
8800         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8801         if HAVE_INITFINI_ARRAY_SUPPORT.
8803 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8805         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8806         default search path for VxWorks < 7.
8808 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8810         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8811         (get_format_string): Refer to c_getstr.
8813 2018-08-21  Tom de Vries  <tdevries@suse.de>
8815         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8816         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8817         (debuginfo_early_stop): Declare.
8818         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8819         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8820         (debuginfo_early_stop): New function.
8821         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8822         and debuginfo_early_stop.
8823         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8824         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8825         (general_init): Call debuginfo_early_init.
8826         (finalize): Call debuginfo_fini.
8827         (do_compile): Call debuginfo_init.
8828         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8829         -fdump-early-debug.
8830         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8832 2018-08-21  Tom de Vries  <tdevries@suse.de>
8834         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8835         flag_dump_noaddr and flag_dump_unnumbered.
8837 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8839         * wide-int-range.cc (wide_int_range_abs): New.
8840         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8841         * wide-int-range.h (wide_int_range_abs): New.
8842         (wide_int_range_min_max): New.
8843         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8844         case to call wide_int_range_abs.
8845         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8846         (extract_range_from_abs_expr): Delete.
8848 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8850         PR target/87033
8851         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8852         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8853         for indexed loads.
8855 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8856             Jeff Law <law@redhat.com>
8858         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8859         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8861 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8863         PR other/84889
8864         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8865         (decl_attributes): Likewise.
8866         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8867         instance.
8868         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8869         * diagnostic-core.h (class auto_diagnostic_group): New class.
8870         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8871         (diagnostic_report_diagnostic): Handle the first diagnostics within
8872         a group.
8873         (emit_diagnostic): Add auto_diagnostic_group instance.
8874         (inform): Likewise.
8875         (inform_n): Likewise.
8876         (warning): Likewise.
8877         (warning_at): Likewise.
8878         (warning_n): Likewise.
8879         (pedwarn): Likewise.
8880         (permerror): Likewise.
8881         (error): Likewise.
8882         (error_n): Likewise.
8883         (error_at): Likewise.
8884         (sorry): Likewise.
8885         (fatal_error): Likewise.
8886         (internal_error): Likewise.
8887         (internal_error_no_backtrace): Likewise.
8888         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8889         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8890         * diagnostic.h (struct diagnostic_context): Add fields
8891         "diagnostic_group_nesting_depth",
8892         "diagnostic_group_emission_count", "begin_group_cb",
8893         "end_group_cb".
8894         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8895         Add auto_diagnostic_group instance(s).
8896         (find_explicit_erroneous_behavior): Likewise.
8897         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8898         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8899         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8900         (gimplify_va_arg_expr): Likewise.
8901         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8902         (HSA_SORRY_AT): Likewise.
8903         * ipa-devirt.c (compare_virtual_tables): Likewise.
8904         (warn_odr): Likewise.
8905         * multiple_target.c (expand_target_clones): Likewise.
8906         * opts-common.c (cmdline_handle_error): Likewise.
8907         * reginfo.c (globalize_reg): Likewise.
8908         * substring-locations.c (format_warning_n_va): Likewise.
8909         * tree-inline.c (expand_call_inline): Likewise.
8910         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8911         * tree-ssa-loop-niter.c
8912         (do_warn_aggressive_loop_optimizations): Likewise.
8913         * tree-ssa-uninit.c (warn_uninit): Likewise.
8914         * tree.c (warn_deprecated_use): Likewise.
8916 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
8918         PR target/87014
8919         * config/i386/i386.md (eh_return): Always update EH return
8920         address in word_mode.
8922 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
8924         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8925         TARGET_SPLIT_COMPLEX_ARG is defined.
8927 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8929         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8931 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8933         PR target/86984
8934         * expr.c (expand_assignment): Assert that bitpos is positive.
8935         (store_field): Likewise
8936         (expand_expr_real_1): Make sure that bitpos is positive.
8937         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8938         integer overflow.
8940 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8942         * Makefile.in (CPP_ID_DATA_H): Delete.
8943         (CPP_INTERNAL_H): Don't add it.
8944         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8945         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8947 2018-08-20  Richard Biener  <rguenther@suse.de>
8949         PR tree-optimization/78655
8950         * tree-vrp.c (extract_range_from_binary_expr_1): Make
8951         pointer + offset nonnull if either operand is nonnull work.
8953 2018-08-20  Tom de Vries  <tdevries@suse.de>
8955         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8956         unless the referenced die describes the added property using
8957         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
8958         Otherwise, add a DW_AT_location to the referenced die.
8960 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
8962         PR target/86994
8963         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8964         register_operand when calling ix86_set_reg_reg_cost.
8965         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8966         Set *total to 0 for operands that satisfy x86_64_immediate_operand
8967         predicate and to 1 otherwise.
8969 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
8971         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8972         emit a diagnostic that it is not supported and reset the option.
8973         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8974         supported and consume the option.  (ASM_FINAL_SPEC): New.
8976 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
8978         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8979         a sentence.
8981 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
8983         C-SKY port: Documentation
8985         * doc/extend.texi (C-SKY Function Attributes): New section.
8986         * doc/invoke.texi (Option Summary): Add C-SKY options.
8987         (C-SKY Options): New section.
8988         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8990 2018-08-17  Jojo  <jijie_rong@c-sky.com>
8991             Huibin Wang  <huibin_wang@c-sky.com>
8992             Sandra Loosemore  <sandra@codesourcery.com>
8993             Chung-Lin Tang  <cltang@codesourcery.com>
8995         C-SKY port: Backend implementation
8997         * config/csky/*: New.
8998         * common/config/csky/*: New.
9000 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9001             Huibin Wang  <huibin_wang@c-sky.com>
9002             Sandra Loosemore  <sandra@codesourcery.com>
9003             Chung-Lin Tang  <cltang@codesourcery.com>
9004             Andrew Jenner  <andrew@codesourcery.com>
9006         C-SKY port: Configury
9008         * config.gcc (csky-*-*): New.
9009         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9010         * configure: Regenerated.
9012 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9014         * dump-context.h: Include "dumpfile.h".
9015         (dump_context::dump_printf_va): Convert final param from va_list
9016         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9017         ATTRIBUTE_GCC_DUMP_PRINTF.
9018         (dump_context::dump_printf_loc_va): Likewise.
9019         * dumpfile.c: Include "stringpool.h".
9020         (make_item_for_dump_printf_va): Delete.
9021         (make_item_for_dump_printf): Delete.
9022         (class dump_pretty_printer): New class.
9023         (dump_pretty_printer::dump_pretty_printer): New ctor.
9024         (dump_pretty_printer::emit_items): New member function.
9025         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9026         function.
9027         (dump_pretty_printer::emit_item): New member function.
9028         (dump_pretty_printer::stash_item): New member function.
9029         (dump_pretty_printer::format_decoder_cb): New member function.
9030         (dump_pretty_printer::decode_format): New member function.
9031         (dump_context::dump_printf_va): Reimplement in terms of
9032         dump_pretty_printer.
9033         (dump_context::dump_printf_loc_va): Convert final param from va_list
9034         to va_list *.
9035         (dump_context::begin_scope): Reimplement call to
9036         make_item_for_dump_printf.
9037         (dump_printf): Update for change to dump_printf_va.
9038         (dump_printf_loc): Likewise.
9039         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9040         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9041         with %T, %E, and %G.
9042         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9043         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9044         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9045         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9046         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9047         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9048         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9049         within a dump_printf_loc call to "%wu".
9050         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9051         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9052         missing space after "=".
9053         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9054         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9055         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9056         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9057         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9058         duplicate "vectorized" from message.
9060 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9062         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9063         polyNxK_t element's TYPE_STRING_FLAG.
9065 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9067         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9068         (they were unnamed before).  Fix comments.
9070 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9072         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9074 2018-08-17  Richard Biener  <rguenther@suse.de>
9076         PR tree-optimization/86841
9077         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9079 2018-08-17  Martin Liska  <mliska@suse.cz>
9081         * common.opt: Remove Warn, Init and Report for options with
9082         Ignore/Deprecated flag. Warning is done automatically for
9083         Deprecated flags.
9084         * config/i386/i386.opt: Likewise.
9085         * config/ia64/ia64.opt: Likewise.
9086         * config/rs6000/rs6000.opt: Likewise.
9087         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9088         Remove usage of flag_check_pointer_bounds.
9089         * lto-wrapper.c (merge_and_complain): Do not handle
9090         OPT_fcheck_pointer_bounds.
9091         (append_compiler_options): Likewise.
9092         * opt-functions.awk: Do not handle Deprecated.
9093         * optc-gen.awk: Check that Var, Report and Init are not
9094         used for an option with Ignore/Deprecated flag.
9095         * opts-common.c (decode_cmdline_option): Do not report
9096         CL_ERR_DEPRECATED.
9097         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9098         options.
9099         * opts.h (struct cl_option): Remove cl_deprecated flag.
9100         (CL_ERR_DEPRECATED): Remove error enum value.
9102 2018-08-17  Richard Biener  <rguenther@suse.de>
9104         PR middle-end/86505
9105         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9106         across a va-arg-pack using call adjust its return value accordingly.
9108 2018-08-16  Martin Sebor  <msebor@redhat.com>
9110         PR tree-optimization/86853
9111         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9112         (struct fmtresult): Add member and initialize it in ctors.
9113         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9114         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9115         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9116         (parse_directive): Handle %C same as %c.
9117         (sprintf_dom_walker::compute_format_length): Adjust.
9118         (is_call_safe): Adjust.
9120 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9122         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9123         for determining how to count the elements.
9124         * builtins.h (c_strlen): Adjust prototype.
9125         * expr.c (string_constant): Add new parameter mem_size.
9126         Set *mem_size appropriately.
9127         * expr.h (string_constant): Adjust protoype.
9128         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9129         * gimple-fold.h (get_range_strlen): Adjust prototype.
9130         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9131         (format_string): Call get_string_length with eltsize.
9133 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9135         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9136         to emit the span, rather than setting it as the prefix.
9138 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9140         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9141         "margin_char" parameter, defaulting to space.  Use it in place
9142         of pp_space for the initial part of the margin.
9143         (layout::print_leading_fixits): Use '+' when filling the margin
9144         of line-insertion fix-it hints.
9146 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9148         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9149         Delete.
9151 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9153         * config/rs6000/altivec.md: Don't set length attribute to the default
9154         value.
9155         * config/rs6000/darwin.md: Ditto.
9156         * config/rs6000/dfp.md: Ditto.
9157         * config/rs6000/htm.md: Ditto.
9158         * config/rs6000/rs6000.md: Ditto.
9159         * config/rs6000/sync.md: Ditto.
9160         * config/rs6000/vsx.md: Ditto.
9162 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9164         * config/rs6000/altivec.md: Don't set length attribute to the default
9165         value, for branch instructions.
9166         * config/rs6000/darwin.md: Ditto.
9167         * config/rs6000/rs6000.md: Ditto.
9169 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9171         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9172         (unnamed conditional branch define_insn): Set length to 4 or 8
9173         depending on offset.
9174         (<bd>_<mode>): Similar, for alternative 0.
9175         (<bd>tf_<mode>): Ditto.
9177 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9179         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9181 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9183         * doc/rtl.texi: Replace old RTX class names with new names.
9186 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9188         * expmed.h (canonicalize_comparison): New declaration.
9189         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9190         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9191         * optabs.c (prepare_cmp_insn): Likewise.
9192         * rtl.h (unsigned_condition_p): New function which checks if a
9193         comparison operator is unsigned.
9195 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9197         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9198         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9200 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9202         PR target/84711
9203         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9204         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9205         (mov<mov>): ..this and enable unconditionally.
9207 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9209         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9211 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9213         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9214         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9215         "Common" with "Target".
9217 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9219         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9220         * doc/invoke.texi (mmitigate-rop): Remove.
9221         * config/i386/i386.c: Do not include "regrename.h".
9222         (ix86_rop_should_change_byte_p, reg_encoded_number)
9223         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9224         Remove.
9225         (ix86_reorg): Remove call to ix86_mitigate_rop.
9226         * config/i386/i386.md (attr "modrm_class"): Remove.
9227         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9228         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9229         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9231 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9233         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9234         allow folding of mergeh() and mergel() for the float and double types.
9235         (fold_mergehl_helper): Rework to handle building a permute tree
9236         for float vectors.
9238 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9240         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9241         for TARGET_SSE.
9243 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9245         * common.opt (fdiagnostics-show-labels): New option.
9246         * diagnostic-show-locus.c (class layout_range): Add field
9247         "m_label".
9248         (class layout): Add field "m_show_labels_p".
9249         (layout_range::layout_range): Add param "label" and use it to
9250         initialize m_label.
9251         (make_range): Pass in NULL for new "label" param of layout_range's
9252         ctor.
9253         (layout::layout): Initialize m_show_labels_p.
9254         (layout::maybe_add_location_range): Pass in loc_range->m_label
9255         when constructing layout_range instances.
9256         (struct line_label): New struct.
9257         (layout::print_any_labels): New member function.
9258         (layout::print_line): Call it if label-printing is enabled.
9259         (selftest::test_one_liner_labels): New test.
9260         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9261         * diagnostic.c (diagnostic_initialize): Initialize
9262         context->show_labels_p.
9263         * diagnostic.h (struct diagnostic_context): Add field
9264         "show_labels_p".
9265         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9266         -fno-diagnostics-show-labels.
9267         * dwarf2out.c (gen_producer_string): Add
9268         OPT_fdiagnostics_show_labels to the ignored options.
9269         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9270         param.
9271         (gcc_rich_location::maybe_add_expr): Likewise.
9272         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9273         label" param, defaulting to NULL.
9274         (gcc_rich_location::add_expr): Add "label" param.
9275         (gcc_rich_location::maybe_add_expr): Likewise.
9276         (class text_range_label): New class.
9277         (class range_label_for_type_mismatch): New class.
9278         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9279         of format_warning_va.
9280         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9281         * lto-wrapper.c (merge_and_complain): Add
9282         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9283         (append_compiler_options): Likewise to the dropped options.
9284         (append_diag_options): Likewise to the passed-on options.
9285         * opts.c (common_handle_option): Handle the new option.
9286         * selftest-diagnostic.c
9287         (test_diagnostic_context::test_diagnostic_context): Enable
9288         show_labels_p.
9289         * substring-locations.c: Include "gcc-rich-location.h".
9290         (format_warning_n_va): Add "fmt_label" and "param_label" params
9291         and use them as appropriate.
9292         (format_warning_va): Add "fmt_label" and "param_label" params,
9293         passing them on to format_warning_n_va.
9294         (format_warning_at_substring): Likewise.
9295         (format_warning_at_substring_n): Likewise.
9296         * substring-locations.h (format_warning_va): Add "fmt_label" and
9297         "param_label" params.
9298         (format_warning_n_va): Likewise.
9299         (format_warning_at_substring): Likewise.
9300         (format_warning_at_substring_n): Likewise.
9301         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9303 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9305         PR testsuite/86519
9306         * builtins.c (expand_builtin_memcmp): Do not expand the call
9307         when overflow is detected.
9309 2018-08-15  Martin Sebor  <msebor@redhat.com>
9311         PR tree-optimization/71625
9312         * config/aarch64/aarch64-builtins.c
9313         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9315 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9317         * config/s390/s390.c (s390_reorg): Remove loop.
9319 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9321         * config/darwin.c
9322          (darwin_function_switched_text_sections): Delete.
9323         * gcc/config/darwin.h
9324          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9326 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9328         PR target/81685
9329         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9330         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9331         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9333 2018-08-15  Martin Liska  <mliska@suse.cz>
9335         PR tree-optimization/86925
9336         * predict.c (expr_expected_value_1): When taking
9337         later predictor, assign also probability.
9338         Use fold_build2_initializer_loc in order to fold
9339         the expression in -frounding-math.
9341 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9343         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9344         patterns.
9345         (expand_vec_perm_1): Try the new method.
9347 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9349         PR target/86547
9350         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9351         Check whether lra_live_max_point is 0 before dividing.
9353 2018-08-14  Martin Sebor  <msebor@redhat.com>
9355         PR tree-optimization/86650
9356         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9357         (vrp_prop::check_mem_ref): Same.
9359 2018-08-13  Liu Hao <lh_mouse@126.com>
9361         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9362         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9363         and clear it thereafter, as it only works for DBCS.
9365 2018-08-13  Liu Hao <lh_mouse@126.com>
9367         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9368         handle returned by _get_osf_handle().
9370 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9372         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9373         for folding vec_perm.
9375 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9377         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9378         Add support for gimple-folding of vec_pack() and vec_unpack()
9379         intrinsics.
9381 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9383         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9384         vec_xst variants to the list.
9385         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9386         vector loads and stores.
9388 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9390         * config.gcc (rs6000-ibm-aix4.x): Delete.
9391         (rs6000-ibm-aix5.1): Delete.
9392         (rs6000-ibm-aix5.2): Delete.
9393         (rs6000-ibm-aix5.3): Delete.
9394         * config/rs6000/aix43.h: Delete.
9395         * config/rs6000/aix51.h: Delete.
9396         * config/rs6000/aix52.h: Delete.
9397         * config/rs6000/t-aix43: Delete.
9399 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9401         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9402         New function.
9403         (s390_decompose_address): Factor out constant pool ref
9404         decomposition.
9406 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9408         * config/nds32/nds32-predicates.c
9409         (nds32_can_use_bclr_p): Change return type as bool.
9410         (nds32_can_use_bset_p): Ditto.
9411         (nds32_can_use_btgl_p): Ditto.
9412         (nds32_can_use_bitci_p): Ditto.
9413         * config/nds32/nds32-protos.h
9414         (nds32_can_use_bclr_p): Change declaration.
9415         (nds32_can_use_bset_p): Ditto.
9416         (nds32_can_use_btgl_p): Ditto.
9417         (nds32_can_use_bitci_p): Ditto.
9419 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9421         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9422         Support -msched-prolog-epilog option.
9423         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9425 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9427         * common/config/nds32/nds32-common.c
9428         (nds32_option_optimization_table): Enalbe -malways-align.
9430 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9432         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9433         extra_headers.
9434         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9435         OPT_misr_secure_ case.
9436         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9437         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9438         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9439         secure attribute.
9440         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9441         (nds32_isr_info): New field security_level.
9442         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9443         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9444         * config/nds32/nds32.opt (misr-secure): New option.
9445         * config/nds32/nds32_init.inc: New file.
9446         * config/nds32/nds32_isr.h: New file.
9448 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9450         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9451         Update comment for atomic instructions.
9452         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9453         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9454         Remove.
9455         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9456         load for __ATOMIC_SEQ_CST model.
9457         (atomic_loaddi_1): Remove float register target.
9458         (atomic_storedi): Handle CONST_INT values.
9459         (atomic_storedi_1): Remove float register source.  Add special case
9460         for zero value.
9461         (memory_barrier): New expander and insn.
9463 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9465         PR tree-optimization/86835
9466         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9467         new_stmt after def_gsi, make sure to insert new_square_stmt after
9468         that stmt, not 2 stmts before it.
9470 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9472         PR target/82418
9473         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9474         instead of SWI48.
9476 2018-08-10  Martin Liska  <mliska@suse.cz>
9478         PR target/83610
9479         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9480         function type.
9481         * builtins.c (expand_builtin_expect_with_probability):
9482         New function.
9483         (expand_builtin_expect_with_probability): New function.
9484         (build_builtin_expect_predicate): Add new argumnet probability
9485         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9486         (fold_builtin_expect):
9487         (fold_builtin_2):
9488         (fold_builtin_3):
9489         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9490         * builtins.h (fold_builtin_expect): Set new argument.
9491         * doc/extend.texi: Document __builtin_expect_with_probability.
9492         * doc/invoke.texi: Likewise.
9493         * gimple-fold.c (gimple_fold_call): Pass new argument.
9494         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9495         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9496         * predict.c (get_predictor_value): New function.
9497         (expr_expected_value): Add new argument probability. Assume
9498         that predictor and probability are always non-null.
9499         (expr_expected_value_1): Likewise.  For __builtin_expect and
9500         __builtin_expect_with_probability set probability.  Handle
9501         combination in binary expressions.
9502         (tree_predict_by_opcode): Simplify code by simply calling
9503         get_predictor_value.
9504         (pass_strip_predict_hints::execute): Add handling of
9505         BUILT_IN_EXPECT_WITH_PROBABILITY.
9506         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9507         new predictor.
9508         * tree.h (DECL_BUILT_IN_P): New function.
9510 2018-08-10  Martin Liska  <mliska@suse.cz>
9512         PR tree-optimization/85799
9513         * passes.def: Add argument for pass_strip_predict_hints.
9514         * predict.c (class pass_strip_predict_hints): Add new argument
9515         early_p.
9516         (strip_predictor_early): New function.
9517         (pass_strip_predict_hints::execute): Call the function to
9518         strip predictors.
9519         (strip_predict_hints): New function.
9520         * predict.def: Fix comment.
9522 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9524         * Makefile.in: Clarify which tm.texi to copy over to assert the
9525         right to grant a GFDL license for all.
9527 2018-08-09  Jeff Law  <law@redhat.com>
9529         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9530         unused variable.
9532 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9534         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9535         prototype.
9537 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9539         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9540         reductions for variable-length vectors.
9542 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9544         PR other/84889
9545         * common.opt (fdiagnostics-show-line-numbers): New option.
9546         * diagnostic-show-locus.c (class layout): Add fields
9547         "m_show_line_numbers_p" and "m_linenum_width";
9548         (num_digits): New function.
9549         (test_num_digits): New function.
9550         (layout::layout): Initialize new fields.  Update m_x_offset
9551         logic to handle any left margin.
9552         (layout::print_source_line): Print line number when requested.
9553         (layout::start_annotation_line): New member function.
9554         (layout::print_annotation_line): Call it.
9555         (layout::print_leading_fixits): Likewise.
9556         (layout::print_trailing_fixits): Likewise.  Update calls to
9557         move_to_column for new parameter.
9558         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9559         it to potentially call start_annotation_line.
9560         (layout::show_ruler): Call start_annotation_line.
9561         (selftest::test_line_numbers_multiline_range): New selftest.
9562         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9563         and selftest::test_line_numbers_multiline_range.
9564         * diagnostic.c (diagnostic_initialize): Initialize
9565         show_line_numbers_p.
9566         * diagnostic.h (struct diagnostic_context): Add field
9567         "show_line_numbers_p".
9568         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9569         -fno-diagnostics-show-line-numbers.
9570         * dwarf2out.c (gen_producer_string): Add
9571         OPT_fdiagnostics_show_line_numbers to the ignored options.
9572         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9573         one setting" options.
9574         (append_compiler_options): Likewise to the dropped options.
9575         (append_diag_options): Likewise to the passed-on options.
9576         * opts.c (common_handle_option): Handle the new option.
9577         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9579 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9581         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9582         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9583         third argument of type "const signed char" to descriptions of
9584         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9585         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9586         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9587         __builtin_bcdsub_ov functions.
9589 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9591         PR tree-optimization/86858
9592         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9593         flow_bb_inside_loop_p calls.
9595 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9597         PR tree-optimization/86871
9598         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9599         instead of gimple_assign_lhs.
9601 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9603         PR target/86887
9604         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9605         register constraint to operand 0.
9606         (add<mode>3_carryinC): Likewise.
9607         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9609 2018-08-09  Martin Liska  <mliska@suse.cz>
9611         PR c/86895
9612         * common.opt: Remove extra line.
9614 2018-08-09  Martin Liska  <mliska@suse.cz>
9616         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9617         at the end of a line, make first letter capital and end up
9618         a sentence with a dot.
9619         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9620         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9621         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9622         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9623         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9624         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9625         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9626         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9627         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9628         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9629         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9630         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9631         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9632         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9633         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9634         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9635         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9636         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9638 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9640         PR target/84332
9641         * config/s390/s390.c (s390_option_override_internal): Reduce the
9642         stack-clash-protection-probe-interval param if it would be too big
9643         for z900.
9645 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9647         PR target/46179
9648         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9649         * config/m68k/m68k.c (handle_move_double): Don't call
9650         m68k_final_prescan_insn.
9651         (m68k_adjust_decorated_operand): Renamed from
9652         m68k_final_prescan_insn, remove first and third operand and
9653         simplify.
9654         (print_operand): Call it.
9655         (print_operand_address): Call it.
9657 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9659         * diagnostic.c (diagnostic_report_current_module): Use
9660         linemap_included_from & linemap_included_from_linemap.
9662 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9664         * config/aarch64/aarch64-cores.def: Add phecda core.
9665         * config/aarch64/aarch64-tune.md: Regenerate.
9666         * doc/invoke.texi: Add phecda core.
9668 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9670         PR target/85295
9671         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9672         definitions.
9673         * config/s390/s390.md ("movti"): Add more alternatives for
9674         constant to GPR copies.
9676 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9678         * config/s390/s390.c: Fix whitespace damage throughout the file.
9679         * config/s390/s390.h: Likewise.
9680         * config/s390/tpf.h: Likewise.
9682 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9684         * config/s390/s390.c (s390_loadrelative_operand_p):
9685         Remove TARGET_CPU_ZARCH usages.
9686         (s390_rtx_costs): Likewise.
9687         (s390_legitimate_constant_p): Likewise.
9688         (s390_cannot_force_const_mem): Likewise.
9689         (legitimate_reload_constant_p): Likewise.
9690         (s390_preferred_reload_class): Likewise.
9691         (legitimize_pic_address): Likewise.
9692         (legitimize_tls_address): Likewise.
9693         (s390_split_branches): Removed.
9694         (s390_add_execute): Removed.
9695         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9696         (s390_mainpool_start): Likewise.
9697         (s390_mainpool_finish): Likewise.
9698         (s390_mainpool_cancel): Removed.
9699         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9700         (s390_chunkify_cancel): Likewise.
9701         (s390_return_addr_rtx): Likewise.
9702         (s390_register_info): Remove split_branches_pending_p uages.
9703         (s390_optimize_register_info): Likewise.
9704         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9705         split_branches_pending_p usages.
9706         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9707         (s390_load_got): Likewise.
9708         (s390_expand_split_stack_prologue): Likewise.
9709         (output_asm_nops): Likewise.
9710         (s390_function_profiler): Likewise.
9711         (s390_emit_call): Likewise.
9712         (s390_conditional_register_usage): Likewise.
9713         (s390_optimize_prologue): Likewise.
9714         (s390_reorg): Remove TARGET_CPU_ZARCH and
9715         split_branches_pending_p usages.
9716         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9717         usages.
9718         (s390_output_indirect_thunk_function): Likewise.
9719         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9720         (TARGET_CPU_ZARCH_P): Removed.
9721         (struct machine_function): Remove split_branches_pending_p.
9722         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9724 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9726         * common/config/s390/s390-common.c (processor_flags_table):
9727         Remove flags.
9728         * config.gcc: Remove with_arch/with_tune support.
9729         * config/s390/2064.md: Remove cpu attribute comparisons.
9730         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9731         Remove MTN.
9732         * config/s390/linux.h (ASM_SPEC):
9733         Remove -march support.
9734         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9735         Use a table to get an arch level.
9736         * config/s390/s390-opts.h (enum processor_type):
9737         Remove enum values.
9738         * config/s390/s390.c
9739         (processor_table): Remove entries, add arch_level values.
9740         (s390_issue_rate): Remove cases.
9741         (s390_option_override): Adjust
9742         s390_option_override_internal() call.
9743         (s390_option_override_internal): Remove deprecation warning.
9744         (s390_valid_target_attribute_tree): Adjust
9745         s390_option_override_internal() call.
9746         * config/s390/s390.h (struct s390_processor):
9747         Share with s390-c.c, add arch_level field.
9748         * config/s390/s390.md:
9749         Remove occurrences in cpu attribute.
9750         * config/s390/s390.opt: Remove -march/-mtune support.
9751         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9752         * doc/invoke.texi: Remove deprecation warning.
9754 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9756         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9757         global.
9758         (qdf24xx_tunings): Set vector cost structure to
9759         qdf24xx_vector_cost.
9761         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9762         <register_sextend>: Set to 3.
9764 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9766         PR target/86838
9767         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9768         * config/aarch64/aarch64-simd.md
9769         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9770         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9771         (aarch64_frecpx<mode>): ...this new pattern.
9772         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9773         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9775 2018-08-07  Martin Liska  <mliska@suse.cz>
9777         PR middle-end/83023
9778         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9779         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9780         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9781         * doc/extend.texi: Document that malloc attribute adds
9782         hit to compiler.
9784 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9786         PR target/86785
9787         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9788         Define to speculation_safe_value_not_needed.
9790 2018-08-06  Jeff Law  <law@redhat.com>
9792         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9793         the vr_values instance to cprop_into_stmt.
9794         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9795         (cprop_operand): Also query EVRP to determine if OP is a constant.
9797 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9799         * diagnostic.c (diagnostic_report_current_module): Reroll
9800         included-at loop.  Translate text.
9802 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9804         * function-tests.c (selftest::test_expansion_to_rtl): Call
9805         free_after_compilation.
9807 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9809         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9811 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9813         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9814         loops with memory block operations from getting unrolled.
9816 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9818         PR target/86807
9819         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9820         Define to speculation_safe_value_not_needed.
9822 2018-08-06  Jeff Law  <law@redhat.com>
9824         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9825         assert.
9827 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9829         PR target/86662
9830         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9831         with all enabled __intN types.
9833         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9835 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9837         * alias.c (record_set): Check for clobber high.
9838         * cfgexpand.c (expand_gimple_stmt): Likewise.
9839         * combine-stack-adj.c (single_set_for_csa): Likewise.
9840         * combine.c (find_single_use_1): Likewise.
9841         (set_nonzero_bits_and_sign_copies): Likewise.
9842         (get_combine_src_dest): Likewise.
9843         (is_parallel_of_n_reg_sets): Likewise.
9844         (try_combine): Likewise.
9845         (record_dead_and_set_regs_1): Likewise.
9846         (reg_dead_at_p_1): Likewise.
9847         (reg_dead_at_p): Likewise.
9848         * dce.c (deletable_insn_p): Likewise.
9849         (mark_nonreg_stores_1): Likewise.
9850         (mark_nonreg_stores_2): Likewise.
9851         * df-scan.c (df_find_hard_reg_defs): Likewise.
9852         (df_uses_record): Likewise.
9853         (df_get_call_refs): Likewise.
9854         * dwarf2out.c (mem_loc_descriptor): Likewise.
9855         * haifa-sched.c (haifa_classify_rtx): Likewise.
9856         * ira-build.c (create_insn_allocnos): Likewise.
9857         * ira-costs.c (scan_one_insn): Likewise.
9858         * ira.c (equiv_init_movable_p): Likewise.
9859         (rtx_moveable_p): Likewise.
9860         (interesting_dest_for_shprep): Likewise.
9861         * jump.c (mark_jump_label_1): Likewise.
9862         * postreload-gcse.c (record_opr_changes): Likewise.
9863         * postreload.c (reload_cse_simplify): Likewise.
9864         (struct reg_use): Add source expr.
9865         (reload_combine): Check for clobber high.
9866         (reload_combine_note_use): Likewise.
9867         (reload_cse_move2add): Likewise.
9868         (move2add_note_store): Likewise.
9869         * print-rtl.c (print_pattern): Likewise.
9870         * recog.c (decode_asm_operands): Likewise.
9871         (store_data_bypass_p): Likewise.
9872         (if_test_bypass_p): Likewise.
9873         * regcprop.c (kill_clobbered_value): Likewise.
9874         (kill_set_value): Likewise.
9875         * reginfo.c (reg_scan_mark_refs): Likewise.
9876         * reload1.c (maybe_fix_stack_asms): Likewise.
9877         (eliminate_regs_1): Likewise.
9878         (elimination_effects): Likewise.
9879         (mark_not_eliminable): Likewise.
9880         (scan_paradoxical_subregs): Likewise.
9881         (forget_old_reloads_1): Likewise.
9882         * reorg.c (find_end_label): Likewise.
9883         (try_merge_delay_insns): Likewise.
9884         (redundant_insn): Likewise.
9885         (own_thread_p): Likewise.
9886         (fill_simple_delay_slots): Likewise.
9887         (fill_slots_from_thread): Likewise.
9888         (dbr_schedule): Likewise.
9889         * resource.c (update_live_status): Likewise.
9890         (mark_referenced_resources): Likewise.
9891         (mark_set_resources): Likewise.
9892         * rtl.c (copy_rtx): Likewise.
9893         * rtlanal.c (reg_referenced_p): Likewise.
9894         (single_set_2): Likewise.
9895         (noop_move_p): Likewise.
9896         (note_stores): Likewise.
9897         * sched-deps.c (sched_analyze_reg): Likewise.
9898         (sched_analyze_insn): Likewise.
9900 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9902         * cse.c (invalidate_reg): New function extracted from...
9903         (invalidate): ...here.
9904         (canonicalize_insn): Check for clobber high.
9905         (invalidate_from_clobbers): invalidate clobber highs.
9906         (invalidate_from_sets_and_clobbers): Likewise.
9907         (count_reg_usage): Check for clobber high.
9908         (insn_live_p): Likewise.
9909         * cselib.c (cselib_expand_value_rtx_1):Likewise.
9910         (cselib_invalidate_regno): Check for clobber in setter.
9911         (cselib_invalidate_rtx): Pass through setter.
9912         (cselib_invalidate_rtx_note_stores):
9913         (cselib_process_insn): Check for clobber high.
9914         * cselib.h (cselib_invalidate_rtx): Add operand.
9916 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9918         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9919         (mark_not_eliminable): Likewise.
9920         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9921         * lra-lives.c (process_bb_lives): Check for clobber high.
9922         * lra.c (new_insn_reg): Remember clobber highs.
9923         (collect_non_operand_hard_regs): Check for clobber high.
9924         (lra_set_insn_recog_data): Likewise.
9925         (add_regs_to_insn_regno_info): Likewise.
9926         (lra_update_insn_regno_info): Likewise.
9928 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9930         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9931         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9933 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9935         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9936         (copy_insn_1): Likewise.
9937         (gen_hard_reg_clobber_high): New gen function.
9938         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9939         * genemit.c (gen_exp): Likewise.
9940         (gen_emit_seq): Pass through info.
9941         (gen_insn): Check for CLOBBER_HIGH.
9942         (gen_expand): Pass through info.
9943         (gen_split): Likewise.
9944         (output_add_clobbers): Likewise.
9945         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9946         (remove_clobbers): Likewise.
9947         * rtl.h (gen_hard_reg_clobber_high): New declaration.
9949 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9951         * doc/rtl.texi (clobber_high): Add.
9952         (parallel): Add in clobber high
9953         * rtl.c (rtl_check_failed_code3): Add function.
9954         * rtl.def (CLOBBER_HIGH): Add expression.
9955         * rtl.h (RTL_CHECKC3): Add macro.
9956         (rtl_check_failed_code3): Add declaration.
9957         (XC3EXP): Add macro.
9959 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
9961         PR target/86386
9962         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9963         cfun->machine->max_used_stack_alignment if needed.
9965 2018-08-04  Martin Sebor  <msebor@redhat.com>
9967         PR tree-optimization/86571
9968         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9969         NaN output to 4.
9971 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
9973         PR target/86799
9974         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9975         Define.
9977 2018-08-03  Jeff Law  <law@redhat.com>
9979         PR target/86795
9980         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9981         Define to speculation_safe_value_not_needed.
9983 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
9985         * doc/gcov.texi (-x): Remove duplicate "to".
9986         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9987         (-Wif-not-aligned): Remove duplicate "is".
9988         (-flto): Remove duplicate "the".
9989         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9990         duplicate "v5.00.b".
9991         (MSP430 Options): Remove duplicate "and" from the description
9992         of "-mgprel-sec=regexp".
9993         (x86 Options): Remove duplicate copies of "vmldLog102" and
9994         vmlsLog104 from description of "-mveclibabi=type".
9996 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
9998         * internal-fn.h (first_commutative_argument): Declare.
9999         * internal-fn.c (first_commutative_argument): New function.
10000         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10001         restrictions for pattern statements.  Use first_commutative_argument
10002         to look for commutative operands in calls to internal functions.
10004 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10006         * Makefile.in (wide-int-range.o): New.
10007         * tree-vrp.c: Move all the wide_int_* functions to...
10008         * wide-int-range.cc: ...here.
10009         * tree-vrp.h: Move all the wide_int_* prototypes to...
10010         * wide-int-range.h: ...here.
10012 2018-08-03  Tom de Vries  <tdevries@suse.de>
10014         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10015         UI_NONE.
10016         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10017         * except.c (output_function_exception_table): Do early exit if
10018         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10020 2018-08-03  Martin Liska  <mliska@suse.cz>
10022         * predict.c (dump_prediction): Change to 2 digits
10023         in fraction part.
10025 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10027         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10028         neon_dup_q to...
10029         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10031 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10033         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10034         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10035         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10036         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10037         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10039 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10041         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10042         pp_take_prefix when saving the existing prefix.
10043         * diagnostic.c (diagnostic_append_note): Likewise.
10044         * langhooks.c (lhd_print_error_function): Likewise.
10045         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10046         param's type.  Free the existing prefix.
10047         (pp_take_prefix): New function.
10048         (pretty_printer::pretty_printer): Drop the prefix parameter.
10049         Rename the length parameter to match the comment.
10050         (pretty_printer::~pretty_printer): Free the prefix.
10051         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10052         parameter.
10053         (struct pretty_printer): Drop the "const" from "prefix" field's
10054         type and clarify memory management.
10055         (pp_set_prefix): Drop the "const" from the 2nd param.
10056         (pp_take_prefix): New decl.
10058 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10060         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10061         for word_mode_ok here instead of passing as argument.
10062         (expand_block_compare): Change select_block_compare_mode() call.
10063         (expand_strncmp_gpr_sequence): New function.
10064         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10066 2018-08-02  Jeff Law  <law@redhat.com>
10068         PR target/86790
10069         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10070         Define to speculation_safe_value_not_needed.
10072         PR target/86784
10073         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10074         Define to speculation_safe_value_not_needed.
10076 2018-08-02  Tom de Vries  <tdevries@suse.de>
10078         PR target/86660
10079         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10080         function.  Return UI_TARGET unconditionally.
10081         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10082         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10084 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10086         * genemit.c (print_overload_test): Fix typo.
10088 2018-08-02  Richard Biener  <rguenther@suse.de>
10090         PR tree-optimization/86816
10091         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10092         which checks for value availability before querying it.
10093         (gvn_uses_equal): Use it.
10094         (same_succ_hash): Likewise.
10095         (gimple_equal_p): Likewise.
10097 2018-08-02  Nick Clifton  <nickc@redhat.com>
10099         PR target/86813
10100         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10101         Define to speculation_safe_value_not_needed.
10103         PR target/86810
10104         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10105         Define to speculation_safe_value_not_needed.
10107         PR target/86810
10108         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10109         Define to speculation_safe_value_not_needed.
10111         PR target/86803
10112         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10113         Define to speculation_safe_value_not_needed.
10115         PR target/86797
10116         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10117         Define to speculation_safe_value_not_needed.
10119         PR target/86791
10120         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10121         Define to speculation_safe_value_not_needed.
10123         PR target/86789
10124         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10125         Define to speculation_safe_value_not_needed.
10127         PR target/86787
10128         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10129         Define to speculation_safe_value_not_needed.
10131         PR target/86782
10132         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10133         speculation_safe_value_not_needed.
10135         PR target/86781
10136         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10137         to speculation_safe_value_not_needed.
10139 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10141         * doc/md.texi: Expand the documentation of instruction names
10142         to mention port-local uses.  Document '@' in pattern names.
10143         * read-md.h (overloaded_instance, overloaded_name): New structs.
10144         (mapping): Declare.
10145         (md_reader::handle_overloaded_name): New member function.
10146         (md_reader::get_overloads): Likewise.
10147         (md_reader::m_first_overload): New member variable.
10148         (md_reader::m_next_overload_ptr): Likewise.
10149         (md_reader::m_overloads_htab): Likewise.
10150         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10151         m_next_overload_ptr and m_overloads_htab.
10152         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10153         (get_mode_token, get_code_token, get_int_token): New functions.
10154         (map_attr_string): Add an optional argument that passes back
10155         the associated iterator.
10156         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10157         (md_reader::handle_overloaded_name, add_overload_instance): New
10158         functions.
10159         (apply_iterators): Handle '@' names.  Report an error if '@'
10160         is used without iterators.
10161         (initialize_iterators): Initialize the new iterator_group fields.
10162         * genopinit.c (handle_overloaded_code_for)
10163         (handle_overloaded_gen): New functions.
10164         (main): Use them to print declarations of maybe_code_for_* and
10165         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10166         * genemit.c (print_overload_arguments, print_overload_test)
10167         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10168         (main): Use it to print definitions of maybe_code_for_* and
10169         maybe_gen_* functions.
10170         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10171         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10172         instead of explicit mode checks.
10173         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10174         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10175         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10176         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10177         (aarch64_expand_compare_and_swap): Likewise
10178         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10179         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10180         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10181         (aarch64_constant_pool_reload_icode): Delete.
10182         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10183         instead of aarch64_constant_pool_reload_icode.  Use
10184         code_for_aarch64_reload_mov instead of explicit mode checks.
10185         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10186         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10187         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10188         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10189         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10190         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10191         (aarch64_atomic_load_op_code): Delete.
10192         (aarch64_emit_atomic_load_op): Likewise.
10193         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10194         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10195         instead of aarch64_emit_atomic_load_op.
10196         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10197         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10198         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10199         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10200         character before the pattern name.
10201         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10202         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10203         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10204         (aarch64_frecps<mode>): Likewise.
10205         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10206         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10207         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10208         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10209         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10211 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10213         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10214         Allow HFmode constants if TARGET_FP_F16INST.
10216 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10218         PR target/86014
10219         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10220         No longer check last store for clobber of address register.
10222 2018-08-02  Martin Liska  <mliska@suse.cz>
10224         PR gcov-profile/86817
10225         * gcov.c (process_all_functions): New function.
10226         (main): Call it.
10227         (process_file): Move functions processing to
10228         process_all_functions.
10230 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10232         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10233         "const" to the "gimple *" and "rtx_insn *" parameters.
10234         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10235         Likewise.
10236         (dump_location_t::dump_location_t): Likewise.
10238 2018-08-01  Martin Sebor  <msebor@redhat.com>
10240         PR tree-optimization/86650
10241         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10242         rather than a "gcall *".  Directly pass the data of interest
10243         to percent_K_format, rather than building a temporary CALL_EXPR
10244         to hold it.
10245         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10246         (gimple_fold_builtin_strncat): Adjust.
10247         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10248         gcall* argument with gimple*.
10249         * gimple-ssa-warn-restrict.c (check_call): Same.
10250         (wrestrict_dom_walker::before_dom_children): Same.
10251         (builtin_access::builtin_access): Same.
10252         (check_bounds_or_overlap): Same
10253         (maybe_diag_overlap): Same.
10254         (maybe_diag_offset_bounds): Same.
10255         * tree-diagnostic.c (default_tree_printer): Move usage of
10256         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10257         to this callsite.
10258         * tree-pretty-print.c (percent_K_format): Add argument.
10259         * tree-pretty-print.h: Add argument.
10260         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10261         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10262         (maybe_diag_stxncpy_trunc): Same.
10263         (handle_builtin_stxncpy): Same.
10264         (handle_builtin_strcat): Same.
10266 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10268         * match.pd: Optimise pointer range checks.
10270 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10272         PR tree-optimization/86758
10273         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10274         to remove pattern statements.
10276 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10278         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10279         result of dfs_enumerate_from when constructing stmt_vec_infos,
10280         instead of additionally calling get_loop_body.
10282 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10284         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10285         parameter.
10286         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10287         When creating an iv, assert that the step is not known to be zero.
10288         (vect_setup_realignment): Update call accordingly.
10289         * tree-vect-stmts.c (vectorizable_store): Likewise.
10290         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10292 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10294         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10295         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10296         (vectorizable_reduction): Likewise.
10297         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10298         (vect_detect_hybrid_slp_stmts): Likewise.
10299         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10301 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10303         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10304         (wide_int_set_zero_nonzero_bits): ...this.
10305         (zero_nonzero_bits_from_vr): Rename to...
10306         (vrp_set_zero_nonzero_bits): ...this.
10307         (extract_range_from_multiplicative_op_1): Abstract wide int
10308         code...
10309         (wide_int_range_multiplicative_op): ...here.
10310         (extract_range_from_binary_expr_1): Extract wide int binary
10311         operations into their own functions.
10312         (wide_int_range_lshift): New.
10313         (wide_int_range_can_optimize_bit_op): New.
10314         (wide_int_range_shift_undefined_p): New.
10315         (wide_int_range_bit_xor): New.
10316         (wide_int_range_bit_ior): New.
10317         (wide_int_range_bit_and): New.
10318         (wide_int_range_trunc_mod): New.
10319         (extract_range_into_wide_ints): New.
10320         (vrp_shift_undefined_p): New.
10321         (extract_range_from_multiplicative_op): New.
10322         (vrp_can_optimize_bit_op): New.
10323         * tree-vrp.h (value_range::dump): New.
10324         (wide_int_range_multiplicative_op): New.
10325         (wide_int_range_lshift):New.
10326         (wide_int_range_shift_undefined_p): New.
10327         (wide_int_range_bit_xor): New.
10328         (wide_int_range_bit_ior): New.
10329         (wide_int_range_bit_and): New.
10330         (wide_int_range_trunc_mod): New.
10331         (zero_nonzero_bits_from_bounds): Rename to...
10332         (wide_int_set_zero_nonzero_bits): ...this.
10333         (zero_nonzero_bits_from_vr): Rename to...
10334         (vrp_set_zero_nonzero_bits): ...this.
10335         (range_easy_mask_min_max): Rename to...
10336         (wide_int_range_can_optimize_bit_op): this.
10337         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10338         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10340 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10342         * tree-vectorizer.h (vect_orig_stmt): New function.
10343         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10344         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10345         (vect_create_epilog_for_reduction): Likewise.
10346         (vectorizable_live_operation): Likewise.
10347         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10348         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10349         * tree-vect-stmts.c (vectorizable_call): Likewise.
10350         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10352 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10354         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10355         argument.
10356         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10357         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10358         (vect_transform_loop): Likewise.
10359         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10361 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10363         * tree-vectorizer.h (vect_schedule_slp): Return void.
10364         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10365         (vect_schedule_slp): Likewise.
10367 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10369         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10370         argument.
10371         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10372         instances before the main loop, if any exist.
10374 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10376         PR tree-optimization/86749
10377         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10378         If the lhs is used in a COND_EXPR, check that it is being used
10379         as the "then" or "else" value.
10381 2018-08-01  Tom de Vries  <tdevries@suse.de>
10383         PR target/86800
10384         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10385         speculation_safe_value_not_needed.
10387 2018-08-01  Richard Biener  <rguenther@suse.de>
10389         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10390         as base and offset.
10392 2018-08-01  Martin Liska  <mliska@suse.cz>
10394         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10395         format how successful transformation is dumped.
10396         (gimple_mod_pow2_value_transform): Likewise.
10397         (gimple_mod_subtract_transform): Likewise.
10398         (gimple_stringops_transform): Likewise.
10400 2018-08-01  Martin Liska  <mliska@suse.cz>
10402         PR value-prof/35543
10403         * value-prof.c (interesting_stringop_to_profile_p):
10404         Simplify the code and add BUILT_IN_MEMMOVE.
10405         (gimple_stringops_transform): Likewise.
10407 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10409         * config/aarch64/aarch64-simd.md
10410         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10411         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10412         use GPI iterator instead of SI mode.
10414 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10416         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10417         rs6000_speculation_barrier.
10418         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10419         new barrier pattern name.
10421 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10423         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10424         (speculation_barrier): New insn.
10426 2018-08-01  Richard Biener  <rguenther@suse.de>
10428         PR bootstrap/86724
10429         * graphite.h: Include isl/id.h and isl/space.h to allow build
10430         with ISL 0.20.
10432 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10434         PR target/86651
10435         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10436         mode for COFF targets.
10437         * defaults.h (TARGET_COFF): Define.
10438         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10439         TARGET_COFF): Define.
10440         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10441         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10442         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10444 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10446         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10447         * gimple-streamer-out.c (output_bb): Save it.
10448         * lto-streamer-in.c (input_struct_function_base): Restore
10449         instance discriminator if available.  Create map on demand.
10450         * lto-streamer-out.c (output_struct_function_base): Save it if
10451         available.
10452         * final.c (decl_to_instance_map): Document LTO strategy.
10454 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10455             Olivier Hainque  <hainque@adacore.com>
10457         * debug.h (decl_to_instance_map_t): New type.
10458         (decl_to_instance_map): Declare.
10459         (maybe_create_decl_to_instance_map): New inline function.
10460         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10461         to track basic block discriminators.
10462         (final_start_function_1): Initialize them.
10463         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10464         bb_discriminator.
10465         (decl_to_instance_map): New variable.
10466         (map_decl_to_instance, maybe_set_discriminator): New functions.
10467         (notice_source_line): Set discriminator.
10469 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10471         * targhooks.c (default_have_speculation_safe_value): Add
10472         ATTRIBUTE_UNUSED.
10474 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10476         * dump-context.h: Include "pretty-print.h".
10477         (dump_context::refresh_dumps_are_enabled): New decl.
10478         (dump_context::emit_item): New decl.
10479         (class dump_context): Add fields "m_test_pp" and
10480         "m_test_pp_flags".
10481         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10482         (temp_dump_context::get_dumped_text): New decl.
10483         (class temp_dump_context): Add field "m_pp".
10484         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10485         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10486         m_test_pp.
10487         (set_dump_file): Update for above change.
10488         (set_alt_dump_file): Likewise.
10489         (dump_loc): New overload, taking a pretty_printer *.
10490         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10491         to any test pretty-printer.
10492         (make_item_for_dump_gimple_stmt): New function, adapted from
10493         optinfo::add_gimple_stmt.
10494         (dump_context::dump_gimple_stmt): Call it, and use the result,
10495         eliminating the direct usage of dump_file and alt_dump_file in
10496         favor of indirectly using them via emit_item.
10497         (make_item_for_dump_gimple_expr): New function, adapted from
10498         optinfo::add_gimple_expr.
10499         (dump_context::dump_gimple_expr): Call it, and use the result,
10500         eliminating the direct usage of dump_file and alt_dump_file in
10501         favor of indirectly using them via emit_item.
10502         (make_item_for_dump_generic_expr): New function, adapted from
10503         optinfo::add_tree.
10504         (dump_context::dump_generic_expr): Call it, and use the result,
10505         eliminating the direct usage of dump_file and alt_dump_file in
10506         favor of indirectly using them via emit_item.
10507         (make_item_for_dump_printf_va): New function, adapted from
10508         optinfo::add_printf_va.
10509         (make_item_for_dump_printf): New function.
10510         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10511         and use the result, eliminating the direct usage of dump_file and
10512         alt_dump_file in favor of indirectly using them via emit_item.
10513         (make_item_for_dump_dec): New function.
10514         (dump_context::dump_dec): Call it, and use the result,
10515         eliminating the direct usage of dump_file and alt_dump_file in
10516         favor of indirectly using them via emit_item.
10517         (make_item_for_dump_symtab_node): New function, adapted from
10518         optinfo::add_symtab_node.
10519         (dump_context::dump_symtab_node): Call it, and use the result,
10520         eliminating the direct usage of dump_file and alt_dump_file in
10521         favor of indirectly using them via emit_item.
10522         (dump_context::begin_scope): Reimplement, avoiding direct usage
10523         of dump_file and alt_dump_file in favor of indirectly using them
10524         via emit_item.
10525         (dump_context::emit_item): New member function.
10526         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10527         Set up test pretty-printer on the underlying context.  Call
10528         refresh_dumps_are_enabled.
10529         (temp_dump_context::~temp_dump_context): Call
10530         refresh_dumps_are_enabled.
10531         (temp_dump_context::get_dumped_text): New member function.
10532         (selftest::verify_dumped_text): New function.
10533         (ASSERT_DUMPED_TEXT_EQ): New macro.
10534         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10535         and then without optinfo enabled.  Add uses of
10536         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10537         * dumpfile.h: Update comment for the dump_* API.
10538         * optinfo-emit-json.cc
10539         (selftest::test_building_json_from_dump_calls): Update for new
10540         param for temp_dump_context ctor.
10541         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10542         and "m_owned" field.
10543         (optinfo_item::~optinfo_item): Likewise.
10544         (optinfo::add_item): New member function.
10545         (optinfo::emit): Update comment.
10546         (optinfo::add_string): Delete.
10547         (optinfo::add_printf): Delete.
10548         (optinfo::add_printf_va): Delete.
10549         (optinfo::add_gimple_stmt): Delete.
10550         (optinfo::add_gimple_expr): Delete.
10551         (optinfo::add_tree): Delete.
10552         (optinfo::add_symtab_node): Delete.
10553         (optinfo::add_dec): Delete.
10554         * optinfo.h (class dump_context): New forward decl.
10555         (optinfo::add_item): New decl.
10556         (optinfo::add_string): Delete.
10557         (optinfo::add_printf): Delete.
10558         (optinfo::add_printf_va): Delete.
10559         (optinfo::add_gimple_stmt): Delete.
10560         (optinfo::add_gimple_expr): Delete.
10561         (optinfo::add_tree): Delete.
10562         (optinfo::add_symtab_node): Delete.
10563         (optinfo::add_dec): Delete.
10564         (optinfo::add_poly_int): Delete.
10565         (optinfo_item::optinfo_item): Remove "owned" param.
10566         (class optinfo_item): Remove field "m_owned".
10568 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10570         PR middle-end/86705
10571         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10572         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10573         requested variable alignment.
10574         (expand_one_ssa_partition): Likewise.
10575         (expand_one_var): Likewise.
10577 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10579         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10580         to speculation_safe_value_not_needed.
10582 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10584         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10585         * targhooks.c (speculation_safe_value_not_needed): New function.
10586         * target.def (have_speculation_safe_value): Update documentation.
10587         * doc/tm.texi: Regenerated.
10589 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10591         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10592         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10593         expand.
10594         (despeculate_copy<ALLI:mode>_insn): New insn.
10595         (despeculate_copyti_insn): New insn.
10596         (despeculate_simple<ALLI:mode>): New insn
10597         (despeculate_simpleti): New insn.
10598         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10599         function.
10600         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10601         aarch64_speculation_safe_value.
10602         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10604 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10606         * config/aarch64/aarch64-speculation.cc: New file.
10607         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10608         before pass_reorder_blocks.
10609         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10610         prototype.
10611         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10612         X14 and X15 when tracking speculation.
10613         * config/aarch64/aarch64.md (register name constants): Add
10614         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10615         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10616         (speculation_barrier): New insn attribute.
10617         (cmp<mode>): Allow SP in comparisons.
10618         (speculation_tracker): New insn.
10619         (speculation_barrier): Add speculation_barrier attribute.
10620         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10621         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10622         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10624 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10626         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10627         aarch64_track_speculation is true.
10628         (tb<optab><mode>1): Likewise.
10629         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10630         generate CB[N]Z when tracking speculation.
10631         (aarch64_split_compare_and_swap): Likewise.
10632         (aarch64_split_atomic_op): Likewise.
10634 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10636         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10638 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10640         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10641         (speculation_barrier): New insn.
10643 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10645         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10646         * config/arm/arm.md (speculation_barrier): New expand.
10647         (speculation_barrier_insn): New pattern.
10649 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10651         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10652         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10653         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10654         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10655         list.
10656         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10657         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10658         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10659         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10660         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10661         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10662         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10663         * builtins.c (expand_speculation_safe_value): New function.
10664         (expand_builtin): Call it.
10665         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10666         * doc/extend.texi: Document __builtin_speculation_safe_value.
10667         * doc/md.texi: Document "speculation_barrier" pattern.
10668         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10669         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10670         * doc/tm.texi: Regenerated.
10671         * target.def (have_speculation_safe_value, speculation_safe_value): New
10672         hooks.
10673         * targhooks.c (default_have_speculation_safe_value): New function.
10674         (default_speculation_safe_value): New function.
10675         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10676         (default_speculation_safe_value): Add prototype.
10678 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10680         * dump-context.h (dump_context::dump_loc): New decl.
10681         * dumpfile.c (dump_context::dump_loc): New member function.
10682         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10683         and dump_gimple_stmt.
10684         (dump_context::dump_gimple_expr_loc): Likewise, using
10685         dump_gimple_expr.
10686         (dump_context::dump_generic_expr_loc): Likewise, using
10687         dump_generic_expr.
10688         (dump_context::dump_printf_loc_va): Likewise, using
10689         dump_printf_va.
10690         (dump_context::begin_scope): Explicitly using the global function
10691         "dump_loc", rather than the member function.
10693 2018-07-31  Martin Sebor  <msebor@redhat.com>
10695         PR tree-optimization/86741
10696         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10698 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10700         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10701         depend on whether prefetch instructions will be emitted or not.
10702         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10703         will be emitted or not.
10704         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10705         (TARGET_SETMEM_PFD): New macros.
10707 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10709         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10710         (NULL_STMT_VEC_INFO): Delete.
10711         (stmt_vec_info::operator*): Likewise.
10712         (stmt_vec_info::operator gimple *): Likewise.
10713         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10714         of NULL_STMT_VEC_INFO.
10715         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10716         (vect_reassociating_reduction_p): Likewise.
10717         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10718         (vectorizable_store): Likewise.
10719         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10720         (vec_info::free_stmt_vec_infos): Likewise.
10722 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10724         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10725         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10727 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10729         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10730         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10731         (vec_info::free_stmt_vec_info): New private member functions.
10732         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10733         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10734         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10735         set_stmt_vec_info_vec and free_stmt_vec_infos.
10736         * tree-vect-loop.c (_loop_vec_info): Remove call to
10737         set_stmt_vec_info_vec.
10738         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10739         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10740         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10741         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10742         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10743         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10744         with the clearing in the xcalloc.
10745         (stmt_vec_info_vec): Delete.
10746         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10747         (vectorize_loops): Likewise.
10748         (vec_info::~vec_info): Remove argument from call to
10749         free_stmt_vec_infos.
10750         (vec_info::add_stmt): Remove vinfo argument from call to
10751         new_stmt_vec_info.
10753 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10755         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10756         rather than a gimple stmt.
10757         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10758         information for pattern statements when passed the original
10759         statement; instead wait to be passed the pattern statement itself.
10760         Don't call set_vinfo_for_stmt here.
10761         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10762         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10763         stmt_vec_infos here.
10764         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10765         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10766         stmt_vec_infos entry.
10768 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10770         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10771         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10772         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10773         * tree-vect-stmts.c (vectorizable_call): Likewise.
10774         (vectorizable_simd_clone_call): Likewise.
10776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10778         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10779         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10780         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10781         * tree-vect-loop.c (vect_transform_loop): Likewise.
10782         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10783         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10785 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10787         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10788         (vect_dr_stmt): Delete.
10789         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10790         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10791         of DR_VECT_AUX.
10792         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10793         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10794         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10795         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10796         (vect_analyze_data_refs): Likewise.
10797         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10798         argument.
10799         (vect_find_same_alignment_drs): Likewise.
10800         (vect_slp_analyze_node_dependences): Update calls accordingly.
10801         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10802         instead of DR_VECT_AUX.
10803         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10804         of a vector data references.  Use vec_info::lookup_dr instead of
10805         DR_VECT_AUX.
10806         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10807         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10808         instead of DR_VECT_AUX.
10810 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10812         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10813         dr_vec_info.
10814         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10815         accordingly.
10816         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10817         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10818         (vect_gen_prolog_loop_niters): Likewise.
10820 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10822         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10823         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10824         (vect_known_alignment_in_bytes, vect_dr_behavior)
10825         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10826         instead of data_references.  Update calls to other routines for
10827         which the same change has been made.
10828         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10829         dr_vec_infos instead of stmt_vec_infos.
10830         (vect_analyze_data_ref_dependence): Update call accordingly.
10831         (vect_slp_analyze_data_ref_dependence)
10832         (vect_record_base_alignments): Use DR_VECT_AUX.
10833         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10834         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10835         (vector_alignment_reachable_p, vect_get_data_access_cost)
10836         (vect_peeling_supportable, vect_analyze_group_access_1)
10837         (vect_analyze_group_access, vect_analyze_data_ref_access)
10838         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10839         (vect_compile_time_alias, vect_small_gap_p)
10840         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10841         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10842         instead of data_references.  Update calls to other routines for
10843         which the same change has been made.
10844         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10845         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10846         (vect_slp_analyze_and_verify_node_alignment)
10847         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10848         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10849         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10850         above changes.
10851         (_vect_peel_info::dr): Replace with...
10852         (_vect_peel_info::dr_info): ...this new field.
10853         (vect_peeling_hash_get_most_frequent)
10854         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10855         (vect_peeling_hash_get_lowest_cost):
10856         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10857         routines for which the same change has been made.
10858         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10859         data_reference.
10860         * tree-vect-loop-manip.c (get_misalign_in_elems)
10861         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10862         above changes.
10863         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10864         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10865         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10866         (get_group_load_store_type, get_negative_load_store_type)
10867         (vect_get_data_ptr_increment, vectorizable_store)
10868         (vectorizable_load): Likewise.
10869         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10870         Update calls to other routines for which the same change has been made.
10872 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10874         * tree-vectorizer.h (vec_info::move_dr): New member function.
10875         (dataref_aux): Rename to...
10876         (dr_vec_info): ...this and add "dr" and "stmt" fields.
10877         (_stmt_vec_info::dr_aux): Update accordingly.
10878         (_stmt_vec_info::data_ref_info): Delete.
10879         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10880         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10881         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10882         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10883         of data_ref.
10884         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
10885         (STMT_VINFO_DR_INFO): New macro.
10886         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10887         (set_dr_misalignment): Update after rename of dataref_aux.
10888         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
10889         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10890         initialization of STMT_VINFO_DATA_REF.
10891         * tree-vectorizer.c (vec_info::move_dr): New function.
10892         * tree-vect-patterns.c (vect_recog_bool_pattern)
10893         (vect_recog_mask_conversion_pattern)
10894         (vect_recog_gather_scatter_pattern): Use it.
10895         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10896         the "dr" and "stmt" fields of dr_vec_info instead of
10897         STMT_VINFO_DATA_REF.
10899 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10901         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10902         (is_pattern_stmt_p): Use it.
10903         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10904         on pattern statements.
10906 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10908         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10909         original stmt as a stmt_vec_info rather than a gimple stmt.
10910         (vect_pattern_recog_1): Take the statement directly as a
10911         stmt_vec_info, rather than via a gimple_stmt_iterator.
10912         Update call to vect_mark_pattern_stmts.
10913         (vect_pattern_recog): Update calls accordingly.
10915 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10917         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10918         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10919         a vect_def_type for the first argument.
10920         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10921         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
10922         operand if it isn't defined by a vectorized statement.
10923         (vect_build_gather_load_calls): Remove the mask_dt argument and
10924         update calls to vect_get_vec_def_for_stmt_copy.
10925         (vectorizable_bswap): Likewise the dt argument.
10926         (vectorizable_call): Update calls to vectorizable_bswap and
10927         vect_get_vec_def_for_stmt_copy.
10928         (vectorizable_simd_clone_call, vectorizable_assignment)
10929         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10930         (vectorizable_comparison): Update calls to
10931         vect_get_vec_def_for_stmt_copy.
10932         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
10933         vect_is_simple_use.
10934         (vect_get_loop_based_defs): Remove dt argument and update call
10935         to vect_get_vec_def_for_stmt_copy.
10936         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10937         and vect_get_vec_def_for_stmt_copy.
10938         (vectorizable_load): Update calls to vect_build_gather_load_calls
10939         and vect_get_vec_def_for_stmt_copy.
10940         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10941         (vectorizable_reduction, vectorizable_live_operation): Update calls
10942         to vect_get_vec_def_for_stmt_copy.
10944 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10946         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10947         and gimple stmt arguments with a stmt_vec_info.
10948         (vect_record_base_alignments): Update calls accordingly.
10949         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10950         and gimple stmt arguments with a stmt_vec_info.
10951         (vect_build_slp_tree_1): Remove vinfo argument and update call
10952         to vect_record_max_nunits.
10953         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10954         and vect_record_max_nunits.
10956 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10958         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10959         file and take a stmt_vec_info instead of a gimple stmt.
10960         (supportable_widening_operation, vect_finish_replace_stmt)
10961         (vect_finish_stmt_generation, vect_get_store_rhs)
10962         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10963         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10964         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10965         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10966         (vect_create_data_ref_ptr, bump_vector_ptr)
10967         (vect_permute_store_chain, vect_setup_realignment)
10968         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10969         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10970         (vectorizable_reduction, vectorizable_induction)
10971         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10972         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10973         than gimple stmts as arguments.
10974         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10975         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10976         (can_group_stmts_p, vect_check_gather_scatter)
10977         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10978         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10979         (vect_permute_load_chain, vect_shift_permute_load_chain)
10980         (vect_transform_grouped_load)
10981         (vect_record_grouped_load_vectors): Likewise.
10982         * tree-vect-loop.c (vect_fixup_reduc_chain)
10983         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10984         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10985         (vectorizable_reduction, vectorizable_induction)
10986         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10987         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10988         (vect_get_load_store_mask): Likewise.
10989         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10990         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10991         * tree-vect-stmts.c (vect_mark_relevant)
10992         (is_simple_and_all_uses_invariant)
10993         (exist_non_indexing_operands_for_use_p, process_use)
10994         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10995         (vect_get_vec_def_for_operand, vect_get_vec_defs)
10996         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10997         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10998         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10999         (get_negative_load_store_type, get_load_store_type)
11000         (vect_check_load_store_mask, vect_check_store_rhs)
11001         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11002         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11003         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11004         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11005         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11006         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11007         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11008         (supportable_widening_operation): Likewise.
11010 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11012         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11013         a stmt_vec_info instead of a gcall.
11014         (vect_check_gather_scatter): Update call accordingly.
11015         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11016         of a gphi.
11017         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11018         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11019         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11020         instead of a gimple stmt.
11021         (vect_transform_loop): Update calls accordingly.
11022         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11023         stmt_vec_infos instead of gimple stmts.
11024         (vect_analyze_slp_instance): Update use accordingly.
11025         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11026         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11027         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11028         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11029         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11030         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11031         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11032         down to subroutines.
11034 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11036         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11037         of the worklist from a vector of gimple stmts to a vector of
11038         stmt_vec_infos.
11039         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11040         (vect_mark_stmts_to_be_vectorized): Likewise
11042 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11044         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11045         statement before passing it to vect_analyze_stmt.
11046         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11047         the chain of phi vector definitions.  Track the exit phi via its
11048         stmt_vec_info.
11049         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11050         STMT_VINFO_REDUC_DEF.
11051         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11052         stmt_vec_infos to handle the statement chains.
11053         (vect_get_slp_defs): Record the first statement in the node
11054         using a stmt_vec_info.
11055         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11056         statements here and pass their stmt_vec_info down to subroutines.
11057         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11058         down to vect_finish_stmt_generation.
11059         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11060         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11061         stmt_vec_infos to subroutines.
11062         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11063         chains.
11065 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11067         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11068         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11069         (vect_permute_store_chain, vect_setup_realignment)
11070         (vect_permute_load_chain, vect_shift_permute_load_chain)
11071         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11072         stmts internally, and when passing values to other vectorizer routines.
11073         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11074         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11075         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11076         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11077         (vectorizable_reduction, vectorizable_induction)
11078         (vectorizable_live_operation, vect_transform_loop_stmt)
11079         (vect_transform_loop): Likewise.
11080         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11081         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11082         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11083         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11084         (vect_slp_analyze_node_operations_1): Likewise.
11085         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11086         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11087         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11088         (vect_finish_stmt_generation_1, get_group_load_store_type)
11089         (get_load_store_type, vect_build_gather_load_calls)
11090         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11091         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11092         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11093         (vectorizable_store, vectorizable_load, vectorizable_condition)
11094         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11095         (supportable_widening_operation): Likewise.
11096         (vect_get_vector_types_for_stmt): Likewise.
11097         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11099 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11101         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11102         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11103         (vect_permute_store_chain, vect_permute_load_chain)
11104         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11105         repeated stmt_vec_info lookups.
11106         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11107         (vect_update_ivs_after_vectorizer): Likewise.
11108         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11109         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11110         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11111         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11112         (vect_bb_slp_scalar_cost): Likewise.
11113         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11115 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11117         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11118         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11119         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11120         to get gassigns and gcalls, rather than operating on generc gimple
11121         stmts.
11122         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11123         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11124         (vectorizable_load, vect_analyze_stmt): Likewise.
11125         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11127 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11129         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11130         return stmt_vec_infos rather than gimple stmts.  Do not accept
11131         null arguments.
11132         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11133         of a gimple stmt.
11134         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11135         Update use of get_later_stmt.
11136         (vect_get_constant_vectors): Update call accordingly.
11137         (vect_schedule_slp_instance): Likewise
11138         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11139         (vect_slp_analyze_instance_dependence): Likewise.
11140         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11142 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11144         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11145         (stmt_info_for_cost::stmt_info): ...this new field.
11146         (add_stmt_costs): Update accordingly.
11147         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11148         (vect_get_known_peeling_cost): Likewise.
11149         (vect_estimate_min_profitable_iters): Likewise.
11150         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11152 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11154         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11155         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11156         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11157         accordingly.
11158         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11160 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11162         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11163         a gimple stmt to a stmt_vec_info.
11164         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11166 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11168         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11169         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11170         (_loop_vec_info::reduction_chains): Likewise.
11171         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11172         accordingly.
11173         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11175 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11177         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11178         a gimple stmt to a stmt_vec_info.
11179         (_stmt_vec_info::next_element): Likewise.
11180         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11181         (vect_slp_analyze_and_verify_node_alignment)
11182         (vect_analyze_group_access_1, vect_analyze_group_access)
11183         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11184         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11185         (vect_supportable_dr_alignment): Update accordingly.
11186         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11187         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11188         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11189         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11190         * tree-vect-slp.c (vect_build_slp_tree_1)
11191         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11192         (vect_split_slp_store_group, vect_analyze_slp_instance)
11193         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11194         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11195         (get_group_load_store_type, get_load_store_type)
11196         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11197         (vect_transform_stmt, vect_remove_stores): Likewise.
11199 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11201         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11202         than a gimple stmt.
11203         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11204         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11205         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11206         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11207         (vector_alignment_reachable_p, vect_get_data_access_cost)
11208         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11209         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11210         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11211         (vect_analyze_group_access_1, vect_analyze_group_access)
11212         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11213         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11214         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11215         result of vect_dr_stmt and use the stmt_vec_info instead of
11216         the associated gimple stmt.
11217         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11218         (vect_gen_prolog_loop_niters): Likewise.
11219         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11221 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11223         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11224         to a vec<stmt_vec_info>.
11225         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11226         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11227         vec<stmt_vec_info>.
11228         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11229         to a vec<stmt_vec_info>.
11230         (bst_traits::value_type, bst_traits::value_type): Likewise.
11231         (bst_traits::hash): Update accordingly.
11232         (vect_get_and_check_slp_defs): Change the stmts parameter from
11233         a vec<gimple *> to a vec<stmt_vec_info>.
11234         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11235         (vect_build_slp_tree): Likewise.
11236         (vect_build_slp_tree_2): Likewise.  Update uses of
11237         SLP_TREE_SCALAR_STMTS.
11238         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11239         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11240         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11241         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11242         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11243         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11244         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11245         (vect_get_constant_vectors, vect_get_slp_defs)
11246         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11247         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11248         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11249         instead of gimple stmts.
11250         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11251         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11252         (vect_slp_analyze_instance_dependence): Update uses of
11253         SLP_TREE_SCALAR_STMTS.
11254         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11255         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11256         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11257         (get_initial_defs_for_reduction): Likewise.
11258         (vect_create_epilog_for_reduction): Likewise.
11259         (vectorize_fold_left_reduction): Likewise.
11260         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11261         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11262         (can_vectorize_live_stmts): Likewise.
11264 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11266         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11267         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11268         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11269         than gimple stmts.
11270         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11271         of a gimple stmt.
11272         (gather_scalar_reductions): Update after above interface changes.
11273         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11274         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11275         than gimple stmts.
11276         (vect_force_simple_reduction): Likewise.
11277         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11278         LOOP_VINFO_REDUCTIONS.
11279         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11281 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11283         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11284         a gimple stmt to a stmt_vec_info.
11285         * tree-vect-loop.c (vect_active_double_reduction_p)
11286         (vect_force_simple_reduction, vectorizable_reduction): Update
11287         accordingly.
11289 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11291         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11292         vec<gimple *> to a vec<stmt_vec_info>.
11293         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11294         the reduction_phis argument from a vec<gimple *> to a
11295         vec<stmt_vec_info>.
11296         (vectorizable_reduction): Likewise the phis local variable that
11297         is passed to vect_create_epilog_for_reduction.  Update for new type
11298         of SLP_TREE_VEC_STMTS.
11299         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11300         (vectorizable_live_operation): Likewise.
11301         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11302         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11304 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11306         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11307         a gimple stmt to a stmt_vec_info.
11308         (vectorizable_condition, vectorizable_live_operation)
11309         (vectorizable_reduction, vectorizable_induction): Pass back the
11310         vectorized statement as a stmt_vec_info.
11311         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11312         use of STMT_VINFO_VEC_STMT.
11313         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11314         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11315         as stmt_vec_infos rather than gimple stmts.
11316         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11317         to a stmt_vec_info.
11318         (vectorizable_live_operation): Likewise.
11319         (vectorizable_reduction, vectorizable_induction): Likewise,
11320         updating use of STMT_VINFO_VEC_STMT.
11321         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11322         of STMT_VINFO_VEC_STMT.
11323         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11324         (vectorizable_simd_clone_call, vectorizable_conversion)
11325         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11326         (vectorizable_store, vectorizable_load, vectorizable_condition)
11327         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11328         from a gimple stmt to a stmt_vec_info.
11329         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11330         pointer to a stmt_vec_info to the vectorizable_* routines.
11332 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11334         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11335         a gimple stmt to a stmt_vec_info.
11336         (is_pattern_stmt_p): Update accordingly.
11337         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11338         (vect_record_grouped_load_vectors): Likewise.
11339         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11340         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11341         (vect_model_reduction_cost): Likewise.
11342         (vect_create_epilog_for_reduction): Likewise.
11343         (vectorizable_reduction, vectorizable_induction): Likewise.
11344         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11345         Return the stmt_vec_info for the pattern statement.
11346         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11347         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11348         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11349         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11350         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11351         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11352         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11353         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11355 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11357         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11358         (vect_finish_stmt_generation): Likewise.
11359         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11360         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11361         (vect_build_gather_load_calls): Use the return value of the above
11362         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11363         scopes for the input gimple stmt and wider scopes for the associated
11364         stmt_vec_info.  Use vec_info::lookup_def when setting these
11365         stmt_vec_infos from an SSA_NAME definition.
11366         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11367         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11368         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11369         (vectorizable_store, vectorizable_load, vectorizable_condition)
11370         (vectorizable_comparison): Likewise.
11371         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11372         (vectorizable_reduction): Likewise.
11374 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11376         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11377         stmt_vec_info * parameter before the optional gimple **.
11378         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11379         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11380         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11381         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11382         (vectorizable_live_operation): Likewise.
11383         * tree-vect-patterns.c (type_conversion_p): Likewise.
11384         (vect_look_through_possible_promotion): Likewise.
11385         (vect_recog_rotate_pattern): Likewise.
11386         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11388 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11390         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11391         a typedef to a wrapper class.
11392         (NULL_STMT_VEC_INFO): New macro.
11393         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11394         (stmt_vec_info::operator*): New function.
11395         (stmt_vec_info::operator gimple *): Likewise.
11396         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11397         (add_stmt_costs): Likewise.
11398         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11399         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11400         (vect_get_known_peeling_cost): Likewise.
11401         (vect_estimate_min_profitable_iters): Likewise.
11402         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11403         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11404         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11405         (vectorizable_store, free_stmt_vec_infos): Likewise.
11406         (new_stmt_vec_info): Change return type of xcalloc to
11407         _stmt_vec_info *.
11409 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11411         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11412         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11413         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11414         a single_imm_use-based sequence.
11415         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11417 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11419         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11420         * tree-vectorizer.c (vec_info::lookup_def): New function.
11421         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11422         (vect_widened_op_tree): Likewise.
11423         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11424         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11425         (vectorizable_reduction): Likewise.
11426         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11427         of a gimple *.
11428         (vect_is_slp_reduction): Update calls accordingly.  Use
11429         vec_info::lookup_def.
11430         (vect_is_simple_reduction): Likewise
11431         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11433 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11435         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11436         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11437         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11438         of vinfo_for_stmt.
11439         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11440         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11441         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11442         (vect_is_slp_reduction, vectorizable_induction)
11443         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11444         * tree-vect-patterns.c (vect_init_pattern_stmt):
11445         (vect_determine_min_output_precision_1, vect_determine_precisions)
11446         (vect_pattern_recog): Likewise.
11447         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11448         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11449         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11450         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11451         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11452         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11453         info field from a loop to a loop_vec_info.
11455 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11457         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11458         (vec_info::add_stmt): Declare.
11459         * tree-vectorizer.c (vec_info::add_stmt): New function.
11460         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11461         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11462         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11463         (vectorizable_induction): Likewise.
11464         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11465         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11466         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11467         (vectorizable_load): Likewise.
11468         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11469         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11470         (vect_recog_gather_scatter_pattern): Likewise.
11471         (append_pattern_def_seq): Likewise.  Remove a check that is
11472         performed by add_stmt itself.
11474 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11476         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11477         which make_ssa_name was called with new_stmt before new_stmt
11478         had been created.
11480 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11482         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11483         split out from...
11484         (vect_is_slp_reduction): ...here...
11485         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11486         that are already known to be false.
11488 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11490         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11491         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11492         STMT_VINFO_NUM_SLP_USES when it's true.
11493         (vect_free_slp_instance): Add a final_p parameter and pass it to
11494         vect_free_slp_tree.
11495         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11496         (vect_analyze_slp_instance): Likewise.
11497         (vect_slp_analyze_operations): Likewise.
11498         (vect_slp_analyze_bb_1): Likewise.
11499         * tree-vectorizer.c (vec_info): Likewise.
11500         * tree-vect-loop.c (vect_transform_loop): Likewise.
11502 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11504         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11505         function is not called for second and subsequent members of
11506         a reduction group.
11508 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11510         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11511         cases for nested loops from here to ...
11512         (vect_create_epilog_for_reduction): ...here.  Only call
11513         vect_is_simple_use for inner-loop reductions.
11515 2018-07-31  Martin Liska  <mliska@suse.cz>
11517         PR gcov-profile/85338
11518         PR gcov-profile/85350
11519         PR gcov-profile/85372
11520         * profile.c (struct location_triplet): New.
11521         (struct location_triplet_hash): Likewise.
11522         (output_location): Do not output a BB that
11523         is already recorded for a line.
11524         (branch_prob): Use streamed_locations.
11526 2018-07-31  Martin Liska  <mliska@suse.cz>
11528         PR gcov-profile/85370
11529         * coverage.c (coverage_begin_function): Do not mark target
11530         clones as artificial functions.
11532 2018-07-31  Martin Liska  <mliska@suse.cz>
11534         PR gcov-profile/83813
11535         PR gcov-profile/84758
11536         PR gcov-profile/85217
11537         PR gcov-profile/85332
11538         * profile.c (branch_prob): Do not record GOTO expressions
11539         for GIMPLE statements which locations are already streamed.
11541 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11543         * gcc.c (handle_spec_function): Accept a soft_matched_part
11544         argument, as do_spec_1.  Pass it down to ...
11545         (eval_spec_function): Accept a soft_matched_part argument,
11546         and pass it down to ...
11547         (do_spec_2): Accept a soft_matched_part argument, and pass
11548         it down to do_spec_1.
11549         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11550         (handle_braces): Update call to handle_spec_function.
11551         (driver::set_up_specs): Update calls to do_spec_2.
11552         (compare_debug_dump_opt_spec_function): Likewise.
11553         (compare_debug_self_opt_spec_function): Likewise.
11555 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11557         * common.opt (nolibc): New option.
11558         * doc/invoke.texi (Link Options): Document it.
11559         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11560         * config/alpha/linux.h: Likewise.
11561         * config/arc/elf.h: Likewise.
11562         * config/arm/uclinux-elf.h: Likewise.
11563         * config/arm/unknown-elf.h: Likewise.
11564         * config/avr/avrlibc.h: Likewise.
11565         * config/bfin/bfin.h: Likewise.
11566         * config/bfin/linux.h: Likewise.
11567         * config/bfin/uclinux.h: Likewise.
11568         * config/darwin.h: Likewise.
11569         * config/darwin10.h: Likewise.
11570         * config/darwin12.h: Likewise.
11571         * config/gnu-user.h: Likewise.
11572         * config/lm32/uclinux-elf.h: Likewise.
11573         * config/pa/pa-hpux11.h: Likewise.
11574         * config/pa/pa64-hpux.h: Likewise.
11575         * config/sparc/sparc.h: Likewise.
11577 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11579         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11580         undefined variables.
11582 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11584         PR target/86640
11585         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11586         instead of GEN_INT.
11588 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11590         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11591         terminated string literal.
11593 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11595         PR rtl-optimization/85160
11596         * combine.c (is_just_move): New function.
11597         (try_combine): Allow combining two instructions into two if neither of
11598         the original instructions was a move.
11600 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11602         PR target/86673
11603         * doc/extend.texi (Global Register Variables): Discourage use of type
11604         qualifiers.
11605         (Local Register Variables): Likewise.
11607 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11609         PR tree-optimization/86506
11610         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11612 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11614         PR target/86547
11615         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11616         hard_regno, make sure no insn between `from` and `to` clobbers it.
11618 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11619             Tom de Vries  <tdevries@suse.de>
11621         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11622         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11623         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11624         PTX_DEFAULT_RUNTIME_DIM.
11625         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11627 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11629         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11630         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11631         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11633 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11635         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11636         used.
11637         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11638         predicate to reflect toc fusion has been deleted.
11639         (toc_fusion_mem_raw): Delete, no longer used.
11640         (toc_fusion_mem_wrapped): Likewise.
11641         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11642         fusion mask bit.
11643         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11644         Delete, no longer used.
11645         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11646         meant to be used for toc fusion.
11647         (rs6000_debug_print_mode): Delete toc fusion debugging.
11648         (rs6000_debug_reg_global): Likewise.
11649         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11650         fusion and secondary reload support that were never used.
11651         (rs6000_option_override_internal): Delete TOC fusion, that was only
11652         partially defined, and it did not work unless you also used the
11653         -mcmodel= switch.
11654         (rs6000_legitimate_address_p): Delete TOC fusion support.
11655         (rs6000_opt_masks): Likewise.
11656         (fusion_wrap_memory_address): Delete function, no longer used.
11657         (fusion_split_address); Delete TOC fusion support.
11658         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11659         longer used with toc fusion being deleted.
11660         (TARGET_TOC_FUSION_FP): Likewise.
11661         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11662         UNSPEC.
11663         (toc fusion spliter): Delete TOC fusion support.
11664         (toc_fusionload_<mode>): Likewise.
11665         (toc_fusionload_di): Likewise.
11666         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11667         longer needs to be named.  Rename predicate to delete TOC fusion.
11668         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11669         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11670         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11671         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11672         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11674 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11676         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11677         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11678         __int128 in built-in function prototypes.
11679         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11680         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11682 2018-07-27  Martin Sebor  <msebor@redhat.com>
11684         PR tree-optimization/86696
11685         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11686         types, including enums.
11687         (handle_char_store): Be prepared for the above function to fail.
11689 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11691         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11692         when optimization level is lower than 2 or optimize for size.
11694 2018-07-26  Martin Sebor  <msebor@redhat.com>
11696         PR tree-optimization/86043
11697         PR tree-optimization/86042
11698         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11699         (get_string_cst_length): Rename...
11700         (get_min_string_length): ...to this.  Add argument.
11701         (handle_char_store): Extend to handle multi-character stores by
11702         MEM_REF.
11703         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11704         * tree.h (initializer_zerop): Add argument.
11706 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11708         PR middle-end/86660
11709         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11710         declare target to variables if they have always,{to,from,tofrom} map
11711         kinds.
11713 2018-07-26  Martin Liska  <mliska@suse.cz>
11715         PR lto/86548
11716         * lto-wrapper.c: Add linker_output as prefix
11717         for ltrans_output_file.
11719 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11721         PR rtl-optimization/85805
11722         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11723         value for hard registers if that was written in the same mode.
11725 2018-07-26  Martin Liska  <mliska@suse.cz>
11727         PR gcov-profile/86536
11728         * gcov.c (format_gcov): Use printf format %.*f directly
11729         and do not handle special values.
11731 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11733         * common/config/arc/arc-common.c (arc_option_optimization_table):
11734         Update default optimizations for size.
11736 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11738         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11739         * config/arc/constraints.md (CfZ): New constraint.
11740         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11741         (subsf3_fpu): Likewise.
11742         (cmpsf_fpu): Likewise.
11743         (cmpsf_fpu_uneq): Likewise.
11745 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11747         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11748         accesses as well.
11749         (arc_is_uncached_mem_p): uncached applies to both the variable and
11750         the pointer.
11752 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11754         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11755         register names.
11757 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11759         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11760         field "m_scopes" from vec to auto_vec.
11762 2018-07-25  Martin Liska  <mliska@suse.cz>
11764         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11765         return type.
11767 2018-07-25  Richard Biener  <rguenther@suse.de>
11769         PR debug/86654
11770         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11771         special wrt context_die late.
11772         (gen_subprogram_die): Re-use DIEs in local scope.
11774 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11776         PR tree-optimization/86644
11777         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11779 2018-07-25  Martin Liska  <mliska@suse.cz>
11781         PR middle-end/86645
11782         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11783         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11785 2018-07-25  Martin Liska  <mliska@suse.cz>
11787         PR sanitizer/79635
11788         * params.def: Explain ASan abbreviation and provide
11789         a documentation link.
11791 2018-07-24  Martin Sebor  <msebor@redhat.com>
11793         PR tree-optimization/86622
11794         PR tree-optimization/86532
11795         * builtins.h (string_length): Declare.
11796         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11797         (check_access): Be prepared for non-constant length ranges.
11798         (string_length): Make extern.
11799         * expr.c (string_constant): Only handle the minor non-constant
11800         array index.  Use string_constant to compute the length of
11801         a generic string constant.
11803 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11805         PR tree-optimization/86618
11806         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11807         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11809 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11811         PR tree-optimization/86636
11812         * json.cc (json::object::set): Fix comment.  Add assertions.
11813         (json::array::append): Move here from json.h.  Add comment and an
11814         assertion.
11815         (json::string::string): Likewise.
11816         * json.h (json::array::append): Move to json.cc.
11817         (json::string::string): Likewise.
11818         * optinfo-emit-json.cc
11819         (optrecord_json_writer::impl_location_to_json): Assert that we
11820         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11821         wrapper around it.  Expand the location once, rather than three
11822         times.
11823         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11824         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11825         wrappers.
11826         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11827         places.  Fix some overlong lines.
11829 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11831         * config/aarch64/aarch64-simd.md
11832         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11833         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11834         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11835         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11836         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11837         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11838         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11839         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11840         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11842 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11844         PR middle-end/86627
11845         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11846         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11847         and abs_d == d, do the power of two handling if profitable.
11849 2018-07-24  Richard Biener  <rguenther@suse.de>
11851         * match.pd: Add BIT_FIELD_REF canonicalizations.
11853 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11855         PR c/86617
11856         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11858 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11860         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11861         terminated STRING_CST object.
11863 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11865         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11867 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11869         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11870         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11871         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11872         the elements into a register.
11873         (rs6000_split_v4si_init_di_reg): Delete.
11874         (rs6000_split_v4si_init): Delete.
11875         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11876         (vsx_init_v4si): Rewrite as a define_expand.
11878 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11880         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11881         zero_extend argument from memory): New.
11883 2018-07-22  Martin Sebor  <msebor@redhat.com>
11885         PR bootstrap/86621
11886         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11887         diagnosing calls with unknown arguments unless -Walloca-larger-than
11888         is restricted to less than PTRDIFF_MAX bytes.
11890 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
11892         * doc/gcov.texi (Invoking Gcov): Editorial changes.
11894 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11896         * pretty-print.c (text_info::set_location): Remove redundant
11897         "line_table" parameter from call to rich_location::set_range.
11899 2018-07-20  Martin Sebor  <msebor@redhat.com>
11901         PR middle-end/82063
11902         * builtins.c (expand_builtin_alloca): Adjust.
11903         * calls.c (alloc_max_size): Simplify.
11904         * cgraphunit.c (cgraph_node::expand): Adjust.
11905         * common.opt (larger_than_size, warn_frame_larger_than): Remove
11906         variables.
11907         (frame_larger_than_size): Same.
11908         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11909         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11910         Initialize.
11911         * doc/invoke.texi (GCC Command Options): Document option arguments.
11912         Explain byte-size arguments and suffixes.
11913         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11914         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11915         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11916         * doc/options.texi (UInteger): Expand.
11917         (Host_Wide_Int, ByteSize): Document new properties.
11918         * final.c (final_start_function_1): Include sizes in an error message.
11919         * function.c (frame_offset_overflow): Same.
11920         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11921         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11922         Diagnose unbounded alloca calls only for limits of less than
11923         PTRDIFF_MAX.
11924         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
11925         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
11926         for alloca(0).
11927         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
11928         only for limits of less than PTRDIFF_MAX.
11929         * langhooks-def.h (lhd_handle_option): Change function argument
11930         to HOST_WIDE_INT.
11931         * langhooks.c (lhd_handle_option): Same.
11932         * langhooks.h (handle_option): Same.
11933         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11934         ByteSize flags.
11935         (var_type, var_type_struct): Same.
11936         (var_set): Handle ByteSize flag.
11937         * optc-gen.awk: Add comments to output to ease debugging.  Make
11938         use of HOST_WIDE_INT where appropriate.
11939         * opts-gen-save.awk:  Use %lx to format unsigned long.
11940         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11941         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11942         arguments.  Parse bytes-size suffixes.
11943         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11944         (enum_value_to_arg): Same.
11945         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
11946         (handle_option): Adjust.
11947         (generate_option): Change function argument to HOST_WIDE_INT.
11948         (cmdline_handle_error): Adjust.
11949         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11950         (set_option): Change function argument to HOST_WIDE_INT.
11951         (option_enabled): Handle cl_host_wide_int.
11952         (get_option_state): Handle CLVC_SIZE.
11953         (control_warning_option): Same.
11954         * opts.c (common_handle_option): Change function argument to
11955         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
11956         OPT_Wvla_larger_than_.
11957         * opts.h (enum cl_var_type): Add an enumerator.
11958         * stor-layout.c (layout_decl): Print a more meaningful warning.
11959         * toplev.c (output_stack_usage): Adjust.
11961 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
11963         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11964         call to inline_expand_builtin_string_cmp.
11965         (expand_builtin_strcmp): Likewise.
11966         (expand_builtin_strncmp): Likewise.
11967         (inline_string_cmp): Delete the last parameter, change char_type_node
11968         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11969         two operands.
11970         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11971         the inlining expansion on target where the type of the call has same or
11972         narrower precision than unsigned char.
11974 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
11976         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11977         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11978         * common.opt (fsave-optimization-record): New option.
11979         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11980         * doc/invoke.texi (-fsave-optimization-record): New option.
11981         * dumpfile.c: Include "optinfo-emit-json.h".
11982         (struct kv_pair): Move to coretypes.h.
11983         (optgroup_options): Make non-static.
11984         (dump_context::end_scope): Call
11985         optimization_records_maybe_pop_dump_scope.
11986         * dumpfile.h (optgroup_options): New decl.
11987         * json.cc: New file.
11988         * json.h: New file.
11989         * optinfo-emit-json.cc: New file.
11990         * optinfo-emit-json.h: New file.
11991         * optinfo.cc: Include "optinfo-emit-json.h".
11992         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11993         (optinfo_enabled_p): Check optimization_records_enabled_p.
11994         (optinfo_wants_inlining_info_p): Likewise.
11995         * optinfo.h: Update comment.
11996         * profile-count.c (profile_quality_as_string): New function.
11997         * profile-count.h (profile_quality_as_string): New decl.
11998         (profile_count::quality): New accessor.
11999         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12000         and optinfo_emit_json_cc_tests.
12001         * selftest.h (selftest::json_cc_tests): New decl.
12002         (selftest::optinfo_emit_json_cc_tests): New decl.
12003         * toplev.c: Include "optinfo-emit-json.h".
12004         (compile_file): Call optimization_records_finish.
12005         (do_compile): Call optimization_records_start.
12006         * tree-ssa-live.c: Include optinfo.h.
12007         (remove_unused_scope_block_p): Retain inlining information if
12008         optinfo_wants_inlining_info_p returns true.
12010 2018-07-20  Richard Biener  <rguenther@suse.de>
12012         PR debug/86585
12013         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12014         to cover -flto-partition=none.
12016 2018-07-20  Martin Liska  <mliska@suse.cz>
12018         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12019         (get_decl_source_range): Remove unused function.
12021 2018-07-20  Richard Biener  <rguenther@suse.de>
12023         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12024         (struct vn_phi_s): Likewise.
12025         (struct vn_reference_s): Likewise.
12026         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12027         for searching the slot of an entry known to be in the hash itself.
12028         (vn_phi_hasher::equal): Likewise.
12029         (vn_reference_hasher::equal): Likewise.
12030         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12031         globals.
12032         (optimistic_info, current_info): Remove, keeping only valid_info.
12033         (vn_reference_lookup_1): Remove fallback lookup.
12034         (vn_reference_lookup_2): Likewise.
12035         (vn_nary_op_lookup_1): Likewise.
12036         (vn_phi_lookup): Likewise.
12037         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12038         hash element.
12039         (vn_reference_insert): Adjust, chain the inserted hash element
12040         at last_inserted_ref.
12041         (vn_reference_insert_pieces): Likewise.
12042         (visit_reference_op_call): Likewise.
12043         (vn_nary_op_insert_into): Chain the inserted hash element at
12044         last_inserted_nary.
12045         (vn_nary_op_insert_pieces): Adjust.
12046         (vn_nary_op_insert): Likewise.
12047         (vn_nary_op_insert_stmt): Likewise.
12048         (vn_phi_insert): Adjust, chain the inserted hash element at
12049         last_inserted_phi.
12050         (process_scc): Remove clearing and copying the optimistic
12051         table.  Instead remove elements inserted during an optimistic
12052         iteration from the single table we maintain.
12053         (init_scc_vn): Adjust.
12054         (free_scc_vn): Likewise.
12055         (sccvn_dom_walker::record_cond): Likewise.
12056         (sccvn_dom_walker::after_dom_children): Likewise.
12058 2018-07-19  Martin Sebor  <msebor@redhat.com>
12060         PR tree-optimization/84047
12061         PR tree-optimization/83776
12062         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12063         (check_array_bounds): Call it.
12065 2018-07-19  Martin Sebor  <msebor@redhat.com>
12067         * align.h (align_flags): Use member initialization.
12069 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12071         * Makefile.in (OBJS): Add optinfo.o.
12072         * coretypes.h (class symtab_node): New forward decl.
12073         (struct cgraph_node): New forward decl.
12074         (class varpool_node): New forward decl.
12075         * dump-context.h: New file.
12076         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12077         "tree-pass.h".
12078         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12079         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12080         (set_alt_dump_file): Likewise.
12081         (dump_context::~dump_context): New dtor.
12082         (dump_gimple_stmt): Move implementation to...
12083         (dump_context::dump_gimple_stmt): ...this new member function.
12084         Add the stmt to any pending optinfo, creating one if need be.
12085         (dump_gimple_stmt_loc): Move implementation to...
12086         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12087         Start a new optinfo and add the stmt to it.
12088         (dump_gimple_expr): Move implementation to...
12089         (dump_context::dump_gimple_expr): ...this new member function.
12090         Add the stmt to any pending optinfo, creating one if need be.
12091         (dump_gimple_expr_loc): Move implementation to...
12092         (dump_context::dump_gimple_expr_loc): ...this new member function.
12093         Start a new optinfo and add the stmt to it.
12094         (dump_generic_expr): Move implementation to...
12095         (dump_context::dump_generic_expr): ...this new member function.
12096         Add the tree to any pending optinfo, creating one if need be.
12097         (dump_generic_expr_loc): Move implementation to...
12098         (dump_context::dump_generic_expr_loc): ...this new member
12099         function.  Add the tree to any pending optinfo, creating one if
12100         need be.
12101         (dump_printf): Move implementation to...
12102         (dump_context::dump_printf_va): ...this new member function.  Add
12103         the text to any pending optinfo, creating one if need be.
12104         (dump_printf_loc): Move implementation to...
12105         (dump_context::dump_printf_loc_va): ...this new member function.
12106         Start a new optinfo and add the stmt to it.
12107         (dump_dec): Move implementation to...
12108         (dump_context::dump_dec): ...this new member function.  Add the
12109         value to any pending optinfo, creating one if need be.
12110         (dump_context::dump_symtab_node): New member function.
12111         (dump_context::get_scope_depth): New member function.
12112         (dump_context::begin_scope): New member function.
12113         (dump_context::end_scope): New member function.
12114         (dump_context::ensure_pending_optinfo): New member function.
12115         (dump_context::begin_next_optinfo): New member function.
12116         (dump_context::end_any_optinfo): New member function.
12117         (dump_context::s_current): New global.
12118         (dump_context::s_default): New global.
12119         (dump_scope_depth): Delete global.
12120         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12121         (dump_symtab_node): New function.
12122         (get_dump_scope_depth): Reimplement in terms of dump_context.
12123         (dump_begin_scope): Likewise.
12124         (dump_end_scope): Likewise.
12125         (selftest::temp_dump_context::temp_dump_context): New ctor.
12126         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12127         (selftest::verify_item): New function.
12128         (ASSERT_IS_TEXT): New macro.
12129         (ASSERT_IS_TREE): New macro.
12130         (ASSERT_IS_GIMPLE): New macro.
12131         (selftest::test_capture_of_dump_calls): New test.
12132         (selftest::dumpfile_c_tests): Call it.
12133         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12134         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12135         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12136         descriptive comment.
12137         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12138         (dump_node, dump_bb): Move these unrelated decls.
12139         (class dump_manager): Add leading comment.
12140         * optinfo.cc: New file.
12141         * optinfo.h: New file.
12143 2018-07-19  Michael Collison  <michael.collison@arm.com>
12144             Richard Henderson <rth@redhat.com>
12146         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12147         (subti): Handle op1 zero.
12148         (subvti4, usub4ti4): New.
12149         (*sub<GPI>3_compare1_imm): New.
12150         (sub<GPI>3_carryinCV): New.
12151         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12152         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12154 2018-07-19  Michael Collison  <michael.collison@arm.com>
12155             Richard Henderson <rth@redhat.com>
12157         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12158         (addti3): Create simpler code if low part is already known to be 0.
12159         (addvti4, uaddvti4): New.
12160         (*add<GPI>3_compareC_cconly_imm): New.
12161         (*add<GPI>3_compareC_cconly): New.
12162         (*add<GPI>3_compareC_imm): New.
12163         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12164         handle constants within this pattern..
12165         (*add<GPI>3_compareV_cconly_imm): New.
12166         (*add<GPI>3_compareV_cconly): New.
12167         (*add<GPI>3_compareV_imm): New.
12168         (add<GPI>3_compareV): New.
12169         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12170         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12171         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12172         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12173         with 'comparison' operator.
12174         (*add<GPI>3_compareV_cconly_imm): Ditto.
12175         (*add<GPI>3_compareV_cconly): Ditto.
12176         (*add<GPI>3_compareV_imm): Ditto.
12177         (add<GPI>3_compareV): Ditto.
12178         (add<mode>3_carryinC): Ditto.
12179         (*add<mode>3_carryinC_zero): Ditto.
12180         (*add<mode>3_carryinC): Ditto.
12181         (add<mode>3_carryinV): Ditto.
12182         (*add<mode>3_carryinV_zero): Ditto.
12183         (*add<mode>3_carryinV): Ditto.
12185 2018-07-19  Michael Collison  <michael.collison@arm.com>
12186             Richard Henderson <rth@redhat.com>
12188         * config/aarch64/aarch64-modes.def (CC_V): New.
12189         * config/aarch64/aarch64-protos.h
12190         (aarch64_addti_scratch_regs): Declare
12191         (aarch64_subvti_scratch_regs): Declare.
12192         (aarch64_expand_subvti): Declare.
12193         (aarch64_gen_unlikely_cbranch): Declare
12194         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12195         for signed overflow using CC_Vmode.
12196         (aarch64_get_condition_code_1): Handle CC_Vmode.
12197         (aarch64_gen_unlikely_cbranch): New function.
12198         (aarch64_addti_scratch_regs): New function.
12199         (aarch64_subvti_scratch_regs): New function.
12200         (aarch64_expand_subvti): New function.
12202 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12204         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12205         extension.
12206         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12207         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12208         extension.
12210 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12212         PR target/83009
12213         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12214         address check not strict.
12216 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12218         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12219         Umq with Umn.
12220         (store_pair_lanes<mode>): Likewise.
12221         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12222         enum value 'ADDR_QUERY_LDP_STP_N'.
12223         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12224         (aarch64_print_address_internal): Add declaration.
12225         (aarch64_print_ldpstp_address): Remove.
12226         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12227         (aarch64_print_operand): Change printing of 'y'.
12228         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12229         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12230         'true' rather than '1'.
12231         * config/aarch64/constraints.md (Uml): Likewise.
12232         (Uml): Rename to Umn.
12233         (Umq): Remove.
12235 2018-07-19  Richard Biener  <rguenther@suse.de>
12237         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12238         a trailing array.
12239         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12240         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12241         (vn_reference_hasher): Likewise.
12242         (struct vn_tables_s): Remove obstack and alloc-pool members.
12243         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12244         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12245         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12246         (vn_reference_insert_pieces): Likewise.
12247         (alloc_vn_nary_op_noinit): Adjust.
12248         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12249         (vn_phi_eq): Adjust.
12250         (shared_lookup_phiargs): Remove.
12251         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12252         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12253         (visit_reference_op_call): Likewise.
12254         (copy_nary, copy_phi, copy_reference): Remove.
12255         (process_scc): Rewind the obstack when iterating.  Do not
12256         copy the elements to valid_info but just move them from one
12257         hashtable to the other.
12258         (allocate_vn_table): Adjust.
12259         (free_vn_table): Likewise.
12260         (init_scc_vn): Likewise.
12261         (free_scc_vn): Likewise.
12263 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12265         PR target/86560
12266         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12267         indirect_return as function type attribute.
12268         (ix86_attribute_table): Change indirect_return to function
12269         type attribute.
12270         * doc/extend.texi: Update indirect_return attribute.
12272 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12274         * wide-int.h (widest2_int): New.
12275         * gimple-fold.c (arith_overflowed_p): Use it.
12276         * tree.h (widest2_int_cst): New.
12277         * tree-vrp.c (wide_int_binop_overflow): Rename from
12278         vrp_int_const_binop.
12279         Rewrite to work on trees.
12280         (extract_range_from_multiplicative_op_1): Abstract code to...
12281         (wide_int_range_min_max): ...here.
12282         (wide_int_range_cross_product): ...and here.
12283         (extract_range_from_binary_expr_1): Abstract overflow code to...
12284         (wide_int_range_mult_wrapping): ...here.
12285         * tree-vrp.h (wide_int_range_cross_product): New.
12286         (wide_int_range_mult_wrapping): New.
12288 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12289             Julia Koval  <julia.koval@intel.com>
12291         * config/i386/x86-tune-costs.h (skylake_memcpy,
12292         skylake_memset): Replace rep_prefix with unrolling for size 512.
12294 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12296         PR middle-end/86544
12297         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12298         comparision with EQ_EXPR in last stmt.
12300 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12302         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12303         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12304         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12305         previously known as "PowerPC AltiVec Built-in Functions".  Move
12306         some material to new subsubsections "PowerPC AltiVec Built-in
12307         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12308         ISA 2.07".
12309         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12310         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12311         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12312         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12314 2018-07-18  Richard Biener  <rguenther@suse.de>
12316         PR tree-optimization/86557
12317         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12318         EXACT_DIV_EXPR.
12320 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12322         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12324 2018-07-17  Jeff Law  <law@redhat.com>
12326         * config/arm/arm.c (get_label_padding): Update for recent
12327         changes to label_to_alignment.
12329         PR tree-optimization/86010
12330         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12332         * config/mips/mips.c (vr4130_align_insns): Update for recent
12333         changes to label_to_alignment.
12335         * config/frv/frv.c (frv_label_align): Update for recent changes
12336         to label_to_alignment.
12338         * config/nios2/nios2.c (nios2_label_align): Update for recent
12339         changes which dropped ALIGN_LABELS_LOG.
12341 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12343         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12344         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12346 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12348         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12349         deprecated align_labels_log.
12351 2018-07-17  Richard Biener  <rguenther@suse.de>
12353         PR lto/86456
12354         * dwarf2out.c (init_sections_and_labels): Always generate
12355         a debug_line_str_section for early LTO debug.
12356         (dwarf2out_finish): Reset debug_line_str_hash output early.
12357         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12358         with fat LTO part.
12359         (dwarf2out_early_finish): Output debug_line_str.
12361 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12363         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12364         index register on z196 or later.
12366 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12368         * config/s390/s390.c (s390_default_align): Set default function
12369         alignment to 16.
12370         (s390_override_options_after_change): Call s390_default align.
12371         (s390_option_override_internal): Call s390_default_align.
12372         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12374 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12376         PR middle-end/86542
12377         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12378         to _looptemp_ clauses, other than the first two.
12380 2018-07-17  Martin Liska  <mliska@suse.cz>
12382         * opts.c: Do not enable OPT_falign_* for -Os.
12384 2018-07-17  Martin Liska  <mliska@suse.cz>
12386         * align.h (MAX_CODE_ALIGN): New.
12387         (MAX_CODE_ALIGN_VALUE): New.
12388         * common/config/i386/i386-common.c (ix86_handle_option):
12389         (MAX_CODE_ALIGN): Moved to align.h.
12390         * final.c (MAX_CODE_ALIGN): Likewise.
12391         * opts.c (parse_and_check_align_values):
12392         (MAX_CODE_ALIGN): Likewise.
12393         (MAX_CODE_ALIGN_VALUE): Likewise.
12395 2018-07-17  Martin Liska  <mliska@suse.cz>
12397         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12398         in order to fulfil coding style.
12399         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12400         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12401         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12402         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12403         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12404         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12405         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12406         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12407         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12409 2018-07-17  Martin Liska  <mliska@suse.cz>
12411         * align.h: New file.
12412         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12413         directly.
12414         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12415         align_flags of label_to_alignment.
12416         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12417         align_flags class.
12418         * config/m68k/m68k.c: Do not use removed align_labels_value and
12419         align_loops_value.
12420         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12421         (LOOP_ALIGN): Likewise.
12422         (LABEL_ALIGN): Likewise.
12423         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12424         Remove not used macro.
12425         (rs6000_loop_align): Change return type to align_flags.
12426         (rs6000_loop_align_max_skip): Remove.
12427         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12428         Change return type to align_flags.
12429         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12430         Remove not used macro.
12431         (rs6000_loop_align):  Change return type to align_flags.
12432         (rs6000_loop_align_max_skip): Remove.
12433         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12434         * config/rx/rx-protos.h (rx_align_for_label): Make it
12435         static function.
12436         * config/rx/rx.c (rx_align_for_label): Change return type
12437         to align_flags.
12438         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12439         macro definitions.
12440         into align_flags class.
12441         (LABEL_ALIGN): Likewise.
12442         (LOOP_ALIGN): Likewise.
12443         * config/s390/s390.c (s390_label_align): Use align_flags
12444         class member.
12445         (s390_asm_output_function_label): Likewise.
12446         * config/sh/sh.c (sh_override_options_after_change):
12447         Use align_flags class directly without macros.
12448         (find_barrier): Likewise.
12449         (barrier_align): Likewise.
12450         (sh_loop_align): Likewise.
12451         * config/spu/spu.c (spu_option_override):
12452         Use align_flags_tuple::get_value instead of removed macros.
12453         (spu_sched_init): Likewise.
12454         * config/spu/spu.h (GTY): Likewise.
12455         * config/visium/visium.c (visium_option_override):
12456         Set "8" as default secondary alignment.
12457         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12458         in order to guarantee secondary alignment of 8.
12459         * coretypes.h: Include align.h header file.
12460         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12461         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12462         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12463         * doc/tm.texi.in: Likewise.
12464         * final.c (struct label_alignment): Remove not used structure.
12465         (LABEL_ALIGN): Change type to align_flags.
12466         (LOOP_ALIGN): Likewise.
12467         (JUMP_ALIGN): Likewise.
12468         (default_loop_align_max_skip): Remove.
12469         (default_label_align_max_skip): Likewise.
12470         (default_jump_align_max_skip): Likewise.
12471         (default_label_align_after_barrier_max_skip):
12472         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12473         (LABEL_TO_MAX_SKIP): Remove.
12474         (label_to_alignment): Return align_flags type instead of integer.
12475         (label_to_max_skip): Remove.
12476         (align_fuzz): Use align_flags type.
12477         (compute_alignments): Use align_flags type and use align_flags::max
12478         to combine multiple alignments.
12479         (grow_label_align): Grow vec instead of C array.
12480         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12481         (shorten_branches):  Use align_flags type and use align_flags::max
12482         to combine multiple alignments.
12483         (final_scan_insn_1): Remove usage of secondary alignment that comes
12484         from label alignment, but instead use proper secondary alignment
12485         which is computed in grow_label_align.
12486         * flags.h (struct align_flags_tuple): Move to align.h.
12487         (struct align_flags): Likewise.
12488         (state_align_loops): Rename to align_loops.
12489         (state_align_jumps): Rename to align_jumps.
12490         (state_align_labels): Rename to align_labels.
12491         (state_align_functions): Rename to align_functions.
12492         (align_loops_log): Remove.
12493         (align_jumps_log): Remove.
12494         (align_labels_log): Remove.
12495         (align_functions_log): Remove.
12496         (align_loops_max_skip): Remove.
12497         (align_jumps_max_skip): Remove.
12498         (align_labels_max_skip): Remove.
12499         (align_functions_max_skip): Remove.
12500         (align_loops_value): Remove.
12501         (align_jumps_value): Remove.
12502         (align_labels_value): Remove.
12503         (align_functions_value): Remove.
12504         * output.h (label_to_alignment): Change return type to align_flags.
12505         (label_to_max_skip): Remove.
12506         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12507         jump_align_max_skip macros.
12508         * targhooks.h (default_loop_align_max_skip): Remove.
12509         (default_label_align_max_skip): Likewise.
12510         (default_jump_align_max_skip): Likewise.
12511         (default_label_align_after_barrier_max_skip): Remove.
12512         * toplev.c (read_log_maxskip): Use ::normalize function.
12513         (parse_N_M): Remove not used argument and also call ::normalize.
12514         (parse_alignment_opts): Do not pass unused arguments.
12515         * varasm.c (assemble_start_function): Use directly align_functions
12516         instead of removed macros.
12517         * system.h: Do not poison removed macros.
12519 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12521         PR middle-end/86539
12522         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12523         and cond temporaries don't have reference type if iterator has
12524         pointer type.  For init use &for_pre_body instead of pre_p if
12525         for_pre_body is non-empty.
12527 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12529         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12530         double-double modes to SFmode directly directly.
12531         (trunc<mode>sf2_fprs): Delete.
12533 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12535         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12536         for conversions between IFmode and the decimal floating point modes.
12537         (init_float128_ieee): Use the correct names for conversions between
12538         KFmode and the decimal floating point modes.
12540 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12542         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12543         for the conversions between TDmode and IFmode.
12544         (init_float128_ieee): Use more correct names for the conversions
12545         between TDmode and KFmode.
12547 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12549         PR tree-optimization/86526
12550         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12551         (inline_expand_builtin_string_cmp): Likewise.
12552         (inline_string_cmp): Likewise.  Use c_readstr instead of
12553         builtin_memcpy_read_str.  Add unit_mode temporary.
12555 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12557         PR middle-end/86528
12558         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12559         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12561 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12563         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12564         Alphabetize prototypes of built-in functions, separating out
12565         built-in functions that are listed in this section but should be
12566         described elsewhere.
12568 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12570         PR target/86511
12571         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12572         conditional move when trapping comparison was split to a
12573         non-trapping one (and vice versa).
12575 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12577         * config/s390/s390.c (s390_function_profiler): Generate nops
12578         instead of profiler call sequences.
12579         * config/s390/s390.opt: Add the new option.
12581 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12583         * config/s390/s390.c (s390_function_profiler): Generate
12584         __mcount_loc section.
12585         * config/s390/s390.opt: Add the new option.
12587 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12589         * common.opt: Add the new warning.
12590         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12591         %r0,__fentry__" when -mfentry is specified.
12592         (s390_option_override_internal): Disallow -mfentry for 31-bit
12593         CPUs.
12594         * config/s390/s390.opt: Add the new option.
12596 2018-07-16  Richard Biener  <rguenther@suse.de>
12598         PR lto/86523
12599         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12600         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12602 2018-07-16  Martin Liska  <mliska@suse.cz>
12604         PR ipa/86529
12605         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12606         to ::get_create.
12608 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12610         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12612 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12614         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12615         for ARCHS4x.
12616         * config/arc/arc-cpus.def (hs4x): New cpu.
12617         (hs4xd): Likewise.
12618         * config/arc/arc-tables.opt: Regenerate.
12619         * config/arc/arc.c (arc_sched_issue_rate): New function.
12620         (TARGET_SCHED_ISSUE_RATE): Define.
12621         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12622         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12623         fpu_cvt.
12624         (attr tune): Add ARCHS4x tune values.
12625         (attr tune_dspmpy): Define.
12626         (*tst): Correct instruction type.
12627         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12628         * config/arc/arcHS4x.md: New file.
12629         * config/arc/fpu.md: Update instruction type attributes.
12630         * config/arc/t-multilib: Regenerate.
12632 2018-07-16  Tom de Vries  <tdevries@suse.de>
12634         PR debug/86455
12635         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12637 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12639         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12640         early clobber.
12642 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12644         PR tree-optimization/86514
12645         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12646         conversion to a boolean type from a type with greater precision.
12648 2018-07-16  Tom de Vries  <tdevries@suse.de>
12650         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12651         dump_flags request TDF_SLIM.
12653 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12655         * fold-const.c (int_const_binop_1): Abstract...
12656         (wide_int_binop): ...wide int code here.
12657         (poly_int_binop): ...poly int code here.
12658         Abstract the rest of int_const_binop_1 into int_const_binop.
12659         * fold-const.h (wide_int_binop): New.
12660         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12661         Remove useless PLUS/MINUS_EXPR case.
12662         (zero_nonzero_bits_from_vr): Move wide int code...
12663         (zero_nonzero_bits_from_bounds): ...here.
12664         (extract_range_from_binary_expr_1): Move mask optimization code...
12665         (range_easy_mask_min_max): ...here.
12666         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12667         (range_easy_mask_min_max): New.
12669 2018-07-15  Jeff Law  <law@redhat.com>
12671         PR target/85993
12672         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12673         block.
12675 2018-07-14  Jim Wilson  <jimw@sifive.com>
12677         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12679 2018-07-14  Paul Koning  <ni1d@arrl.net>
12681         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12683 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12685         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12686         copying section.
12688 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12689             Steve Munroe  <munroesj52@gmail.com>
12691         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12692         (_mm_andnot_si128): Likewise.
12693         (_mm_or_si128): Likewise.
12694         (_mm_xor_si128): Likewise.
12696 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12698         PR middle-end/78809
12699         * builtins.c (expand_builtin_memcmp): Inline the calls first
12700         when result_eq is false.
12701         (expand_builtin_strcmp): Inline the calls first.
12702         (expand_builtin_strncmp): Likewise.
12703         (inline_string_cmp): New routine. Expand a string compare
12704         call by using a sequence of char comparison.
12705         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12706         a call to str(n)cmp/memcmp.
12707         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12708         New option.
12709         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12711 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12713         * config/arm/driver-arm.c: Include arm-native.h.
12714         (host_detect_local_cpu): Use auto-generated data tables.
12715         (vendors, arm_cpu_table): Delete.  Move part information to ...
12716         * config/arm/arm-cpus.in: ... here.
12717         * config/arm/parsecpu.awk (gen_native): New function.
12718         (vendor, part): New CPU fields.
12719         (END): Add support for building the native CPU detection tables.
12720         * config/arm/t-arm (arm-native.h): Add build rule.
12721         (driver-arm.o): Add dependency on arm-native.h.
12723 2018-07-13  Richard Biener  <rguenther@suse.de>
12725         PR middle-end/85974
12726         * match.pd (addr1 - addr2): Allow either of the operand to
12727         have a conversion.
12729 2018-07-13  Tom de Vries  <tdevries@suse.de>
12731         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12732         in remap_ssa_name.
12734 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12736         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12737         arrays instead of numbered variables.
12739 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12741         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12742         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12744 2018-07-13  Richard Biener  <rguenther@suse.de>
12746         PR debug/86452
12747         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12748         instead of get_context_die.
12750 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12751             Richard Biener  <rguenther@suse.de>
12753         PR middle-end/86489
12754         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12755         that the loop latch destination where phi is defined.
12757 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12759         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12760         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12761         (riscv_set_current_function): Extract getting interrupt type to new
12762         function.
12763         (riscv_get_interrupt_type): New function.
12764         (riscv_merge_decl_attributes): New function, checking interrupt type is
12765         same.
12766         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12768 2018-07-12  Paul Koning  <ni1d@arrl.net>
12770         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12771         directive.
12773 2018-07-12  Paul Koning  <ni1d@arrl.net>
12775         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12776         zero reference, add doloop_end instead.
12777         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12778         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12779         detail for doloop_end.
12781 2018-07-12  Martin Sebor  <msebor@redhat.com>
12783         PR c/86453
12784         * attribs.c (decl_attributes): Reject conflicting attributes before
12785         calling attribute handlers.
12787 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12789         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12790          parameter.
12791         (gcc::dump_manager::get_dump_file_name): likewise.
12792         (dump_begin): Likewise.
12793         * dumpfile.h (dump_begin): Update prototype.
12794         (gcc::dump_manager::get_dump_file_name,
12795         gcc::dump_manager::get_dump_file_name): Update prototype.
12797 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12799         * internal-fn.h (vectorizable_internal_fn_p): New function.
12800         * tree-vect-slp.c (compatible_calls_p): Likewise.
12801         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12802         to internal functions.
12803         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12805 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12807         * fold-const.h (inverse_conditions_p): Declare.
12808         * fold-const.c (inverse_conditions_p): New function.
12809         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12810         that test the inverse condition of a conditional internal function.
12811         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12812         * internal-fn.c (internal_fn_mask_index): Handle conditional
12813         internal functions.
12814         (vectorized_internal_fn_supported_p): New function.
12815         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12816         (any_pred_load_store): Replace with...
12817         (need_to_predicate): ...this new variable.
12818         (redundant_ssa_names): New variable.
12819         (ifcvt_can_use_mask_load_store): Move initial checks to...
12820         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12821         for which a conditional internal function exists.
12822         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12823         instead of ifcvt_can_use_mask_load_store.  Update after variable
12824         name change.
12825         (predicate_load_or_store): New function, split out from
12826         predicate_mem_writes.
12827         (check_redundant_cond_expr): New function.
12828         (value_available_p): Likewise.
12829         (predicate_rhs_code): Likewise.
12830         (predicate_mem_writes): Rename to...
12831         (predicate_statements): ...this.  Use predicate_load_or_store
12832         and predicate_rhs_code.
12833         (combine_blocks, tree_if_conversion): Update after above name changes.
12834         (ifcvt_local_dce): Handle redundant_ssa_names.
12835         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12836         general conditional functions.
12837         * tree-vect-stmts.c (vectorizable_call): Likewise.
12839 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12840             Alan Hayward  <alan.hayward@arm.com>
12841             David Sherwood  <david.sherwood@arm.com>
12843         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12844         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12845         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12846         plus and minus and convert them into IFN_COND_FMA-based sequences.
12847         (convert_mult_to_fma): Handle conditional plus and minus.
12849 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12851         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12852         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12853         (cond_fnms_optab): New optabs.
12854         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12855         internal functions.
12856         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12857         * internal-fn.h (get_conditional_internal_fn): Declare.
12858         (get_unconditional_internal_fn): Likewise.
12859         * internal-fn.c (cond_ternary_direct): New macro.
12860         (expand_cond_ternary_optab_fn): Likewise.
12861         (direct_cond_ternary_optab_supported_p): Likewise.
12862         (FOR_EACH_COND_FN_PAIR): Likewise.
12863         (get_conditional_internal_fn): New function.
12864         (get_unconditional_internal_fn): Likewise.
12865         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12866         (gimple_match_op::gimple_match_op): Add a new overload for 5
12867         operands.
12868         (gimple_match_op::set_op): Likewise.
12869         (gimple_resimplify5): Declare.
12870         * genmatch.c (decision_tree::gen): Generate simplifications for
12871         5 operands.
12872         * gimple-match-head.c (gimple_simplify): Define an overload for
12873         5 operands.  Handle calls with 5 arguments in the top-level overload.
12874         (convert_conditional_op): Handle conversions from unconditional
12875         internal functions to conditional ones.
12876         (gimple_resimplify5): New function.
12877         (build_call_internal): Pass a fifth operand.
12878         (maybe_push_res_to_seq): Likewise.
12879         (try_conditional_simplification): Try converting conditional
12880         internal functions to unconditional internal functions.
12881         Handle 3-operand unconditional forms.
12882         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12883         Define ternary equivalents of the current rules for binary conditional
12884         internal functions.
12885         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12886         ternary operations.
12887         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12888         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12889         (optab): Handle them.
12890         (SVE_COND_FP_TERNARY): New int iterator.
12891         (sve_fmla_op, sve_fmad_op): New int attributes.
12892         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12893         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12894         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12896 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12898         * target.def (preferred_else_value): New target hook.
12899         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12900         * doc/tm.texi: Regenerate.
12901         * targhooks.h (default_preferred_else_value): Declare.
12902         * targhooks.c (default_preferred_else_value): New function.
12903         * internal-fn.h (conditional_internal_fn_code): Declare.
12904         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12905         (get_conditional_internal_fn): Use it.
12906         (conditional_internal_fn_code): New function.
12907         * gimple-match.h (gimple_match_cond): New struct.
12908         (gimple_match_op): Add a cond member function.
12909         (gimple_match_op::gimple_match_op): Update all forms to take a
12910         gimple_match_cond.
12911         * genmatch.c (expr::gen_transform): Use the same condition as res_op
12912         for the suboperation, but don't specify a particular else_value.
12913         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12914         (visit_nary_op, visit_reference_op_load): Pass
12915         gimple_match_cond::UNCOND to the gimple_match_op constructor.
12916         * gimple-match-head.c: Include tree-eh.h
12917         (convert_conditional_op): New function.
12918         (maybe_resimplify_conditional_op): Likewise.
12919         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12920         (gimple_resimplify2): Likewise.
12921         (gimple_resimplify3): Likewise.
12922         (gimple_resimplify4): Likewise.
12923         (maybe_push_res_to_seq): Return null for conditional operations.
12924         (try_conditional_simplification): New function.
12925         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
12926         constructor.
12927         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12928         IFN_COND_* call.
12929         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12930         function.
12931         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12933 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12935         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12936         DECL_FCONTEXT
12937         (hash_tree): Do not hash DECL_FCONTEXT
12938         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12939         Do not stream DECL_FCONTEXT.
12940         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12941         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12943 2018-07-12  Richard Biener  <rguenther@suse.de>
12945         PR debug/86462
12946         * dwarf2out.c (gen_block_die): Only output blocks when they have
12947         at least one !DECL_IGNORED_P variable.
12949 2018-07-12  Richard Biener  <rguenther@suse.de>
12951         PR target/84829
12952         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12953         Remove -mieee-fp handling.
12955 2018-07-12  Richard Biener  <rguenther@suse.de>
12957         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12958         left-over from last patch.
12960 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
12962         PR tree-optimization/86492
12963         * gimple-ssa-store-merging.c
12964         (imm_store_chain_info::coalesce_immediate_stores): Call
12965         check_no_overlap even for the merge_overlapping case.  Formatting fix.
12967 2018-07-12  Richard Biener  <rguenther@suse.de>
12969         PR middle-end/86479
12970         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12971         move possibly trapping operations into the conditional.
12973 2018-07-12  Richard Biener  <rguenther@suse.de>
12975         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12976         (vn_lookup_simplify_result): Remove recursion limit applied
12977         here.
12978         (vn_nary_build_or_lookup_1): Adjust.
12979         (try_to_simplify): Likewise.
12980         * gimple-match-head.c (gimple_resimplify1): Instead apply one
12981         here.
12982         (gimple_resimplify2): Likewise.
12983         (gimple_resimplify3): Likewise.
12984         (gimple_resimplify4): Likewise.
12986 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
12988         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12989         Use __mmask64 type instead of __mmask8 for __M argument.
12990         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12991         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12992         __U argument.
12993         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12994         __mmask16 for __M argument.
12995         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12996         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12997         to __mmask16 instead of __mmask8.
12998         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12999         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13000         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13001         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13002         instead of __mmask16 for __U argument.
13003         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13004         __mmask16 instead of __mmask8 for __U argument.
13005         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13006         __U argument.
13007         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13008         __mmask16.
13009         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13010         argument.
13011         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13012         __U argument.
13013         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13014         __mmask16.
13015         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13016         of __mmask16.
13017         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13018         __U argument.
13019         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13020         __U argument.
13021         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13022         __U argument.
13023         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13024         __U argument.
13025         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13026         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13027         return type as well as __M argument type and all casts from __mmask8
13028         to __mmask32.
13029         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13030         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13031         return type as well as __M argument type and all casts from __mmask8
13032         to __mmask16.
13033         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13034         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13035         return type as well as __M argument type and all casts from __mmask8
13036         to __mmask32.
13037         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13038         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13039         return type as well as __M argument type and all casts from __mmask8
13040         to __mmask16.
13041         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13042         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13043         __mmask16.
13045 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13047         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13048         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13049         for __U argument.
13051 2018-07-11  Paul Koning  <ni1d@arrl.net>
13053         * doc/md.texi (define_subst): Document how multiple occurrences of
13054         the same argument in the replacement pattern are handled.
13056 2018-07-11  Paul Koning  <ni1d@arrl.net>
13058         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13059         alphabetical order.
13060         (Common Type Attributes): Add "mode" attribute.
13062 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13064         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13065         stream DECL_ORIGINAL_TYPE.
13066         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13067         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13068         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13069         Do not walk original type.
13070         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13071         external decls.
13072         (write_ts_decl_non_common_tree_pointers): Do not stream
13073         DECL_ORIGINAL_TYPE
13074         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13075         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13077 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13079         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13080         thread twice from the same starting edge.
13082 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13084         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13085         * gimple.c (gimple_call_nonnull_result_p): ...here...
13086         (gimple_call_nonnull_arg): ...and here.
13087         * gimple.h (gimple_call_nonnull_result_p): New.
13088         (gimple_call_nonnull_arg): New.
13090 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13092         * config/arm/arm-cpus.in: Move information from fpu field of each
13093         cpu definition to the isa field.
13094         * config/arm/parsecpu.awk (fpu): Delete match rule.
13095         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13097 2018-07-11  Richard Biener  <rguenther@suse.de>
13099         PR debug/86457
13100         * dwarf2out.c (init_sections_and_labels): Use
13101         output_asm_line_debug_info consistently.
13102         (dwarf2out_early_finish): Likewise.
13103         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13104         type units.
13106 2018-07-11  Richard Biener  <rguenther@suse.de>
13108         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13109         Rework father_bb setting in a way to avoid propagating constants
13110         multiple times on a loop body.
13112 2018-07-10  Mark Wielaard  <mark@klomp.org>
13114         PR debug/86459
13115         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13117 2018-07-10  Richard Biener  <rguenther@suse.de>
13119         * hash-map.h (hash_map::iterator::operator*): Return
13120         references to key and value.
13122 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13124         PR c++/86443
13125         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13126         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13127         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13128         (gimplify_omp_for): For composite loops, move outer
13129         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13130         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13131         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13132         TREE_LIST for both the original class iterator and the "last" helper
13133         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13134         loop, remember has_decl_expr from outer composite loops for the
13135         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13137 2018-07-09  Martin Sebor  <msebor@redhat.com>
13139         PR middle-end/77357
13140         PR middle-end/86428
13141         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13142         accessing implicitly initialized array elements.
13143         * expr.c (string_constant): Handle string initializers of
13144         character arrays within aggregates.
13145         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13146         Store element offset.  As a special case, handle zero size.
13147         (fold_nonarray_ctor_reference): Same.
13148         (fold_ctor_reference): Add argument.  Store subobject offset.
13149         * gimple-fold.h (fold_ctor_reference): Add argument.
13151 2018-07-09  Paul Koning  <ni1d@arrl.net>
13153         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13154         (pdp11_insn_cost): New function.
13155         (pdp11_md_asm_adjust): New function.
13156         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13157         (pdp11_rtx_costs): Update to match machine better.
13158         (output_addr_const_pdp11): Correct format mismatch warnings.
13159         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13160         * config/pdp11/pdp11.md: General change to add base_cost and/or
13161         length attributes for use by new pdp11_insn_cost function.
13162         (MIN_BRANCH): Correct definition.
13163         (MIN_SOB): Ditto.
13164         (doloop_end): Use standard pattern name for looping pattern.
13165         (doloop_end_nocc): New.
13166         (movsf): Add another constraint alternative.
13167         (zero_extendqihi2): Add constraint alternatives for not in place
13168         extend.
13169         (zero_extendhisi2): Remove.
13170         (shift patterns): Add CC handling variants.
13171         (bswaphi2): New.
13172         (bswapsi2): New.
13173         (rothi3): New.
13174         (define_peephole2): New peephole to recognize mov that sets CC for
13175         subsequent test.
13177 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13179         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13180         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13181         wi::add.
13183 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13185         PR c/86420
13186         * real.c (real_nextafter): Return true if result is denormal.
13188 2018-07-09  Martin Liska  <mliska@suse.cz>
13190         * common.opt: Add back wrongly removed attribute.
13192 2018-07-09  Richard Biener  <rguenther@suse.de>
13194         PR debug/86413
13195         * dwarf2out.c (gen_block_die): For an early generated DIE
13196         always output high/low PC attributes.
13198 2018-07-09  Tom de Vries  <tdevries@suse.de>
13200         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13201         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13202         onto VAR_DECL with abstract origin.
13204 2018-07-07  Jim Wilson  <jimw@sifive.com>
13206         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13208 2018-07-07  Tom de Vries  <tdevries@suse.de>
13210         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13212 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13214         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13215         overflow_type.
13216         (combine_bound): Use wide-int overflow calculation instead of
13217         rolling our own.
13218         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13219         overflow_type.
13220         * fold-const.c (int_const_binop_2): Same.
13221         (extract_muldiv_1): Same.
13222         (fold_div_compare): Same.
13223         (fold_abs_const): Same.
13224         * match.pd: Same.
13225         * poly-int.h (add): Same.
13226         (sub): Same.
13227         (neg): Same.
13228         (mul): Same.
13229         * predict.c (predict_iv_comparison): Same.
13230         * profile-count.c (slow_safe_scale_64bit): Same.
13231         * simplify-rtx.c (simplify_const_binary_operation): Same.
13232         * tree-chrec.c (tree_fold_binomial): Same.
13233         * tree-data-ref.c (split_constant_offset_1): Same.
13234         * tree-if-conv.c (idx_within_array_bound): Same.
13235         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13236         * tree-ssa-phiopt.c (minmax_replacement): Same.
13237         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13238         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13239         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13240         * wide-int.cc (wi::add_large): Same.
13241         (wi::mul_internal): Same.
13242         (wi::sub_large): Same.
13243         (wi::divmod_internal): Same.
13244         * wide-int.h: Change overflow type to overflow_type for neg, add,
13245         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13246         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13247         mul_internal, divmod_internal.
13248         (overflow_type): New enum.
13249         (accumulate_overflow): New.
13251 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13253         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13254         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13256 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13258         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13259         argument is checked for zero before entering loop, avoid checking again.
13261 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13263         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13264         * gimplify.h (generic_expr_could_trap_p): Likwise.
13265         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13267 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13269         PR tree-optimization/86401
13270         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13271         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13272         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13273         * fold-const.h (fold_bit_and_mask): Declare.
13274         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13276 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13278         PR target/86324
13279         * target.def (translate_mode_attribute): New hook.
13280         * targhooks.h (default_translate_mode_attribute): Declare.
13281         * targhooks.c (default_translate_mode_attribute): New function.
13282         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13283         * doc/tm.texi: Regenerate.
13284         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13285         (rs6000_translate_mode_attribute): New function.
13287 2018-07-06  Paul Koning  <ni1d@arrl.net>
13289         * doc/md.texi (define_split): Document DONE and FAIL.
13290         (define_peephole2): Ditto.
13292 2018-07-05  Jeff Law  <law@redhat.com>
13294         PR tree-optimization/86010
13295         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13296         both the head and tail of mem* and str* calls.
13298 2018-07-05  Jim Wilson  <jimw@sifive.com>
13300         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13302 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13304         * config/aarch64/aarch64-simd.md: correct flags text for
13305         MIN_EXPR replacement.
13307 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13309         * configure: Regenerated.
13311 2018-07-05  Carl Love  <cel@us.ibm.com>
13313         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13314         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13315         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13316         VSX_BUILTIN_DOUBLEL_V4SF.
13318 2018-07-05  Martin Sebor  <msebor@redhat.com>
13320         PR c++/86400
13321         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13322         than its domain to compute its the upper bound of a char array.
13324 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13326         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13327         * doc/cpp.texi: Update comment.
13328         * doc/tm.texi: Rebuilt.
13329         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13330         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13331         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13332         headers affected by extern "C".
13333         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13334         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13335         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13336         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13337         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13338         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13339         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13340         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13341         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13342         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13343         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13344         config/sparc/sp64-elf.h, config/spu/spu.h,
13345         config/stormy16/stormy16.h, config/v850/v850.h,
13346         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13347         define NO_IMPLICIT_EXTERN_C.
13348         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13350 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13352         PR target/84711
13353         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13354         instead of GET_MODE_SIZE when comparing Units.
13356 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13358         PR target/84711
13359         * rtlanal.c (set_noop_p): Constrain on mode change,
13360         include hard-reg-set.h
13362 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13364         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13366 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13368         Revert
13369         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13371         PR sanitizer/84250
13372         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13373         libasan.
13374         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13376 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13378         PR sanitizer/84250
13379         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13380         libasan.
13381         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13383 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13385         PR middle-end/86380
13386         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13388 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13390         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13391         neg_*_op* variables.
13393 2018-07-04  Martin Liska  <mliska@suse.cz>
13395         * tree-switch-conversion.c: Define
13396         max_ratio_for_speed and max_ratio_for_size constants.
13398 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13399             Martin Liska  <mliska@suse.cz>
13401         PR middle-end/66240
13402         PR target/45996
13403         PR c/84100
13404         * common.opt: Rename align options with 'str_' prefix.
13405         * common/config/i386/i386-common.c (set_malign_value): New
13406         function.
13407         (ix86_handle_option): Use it to set -falign-* options/
13408         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13409         type from int to string.
13410         * config/aarch64/aarch64.c: Update default values from int
13411         to string.
13412         * config/alpha/alpha.c (alpha_override_options_after_change):
13413         Likewise.
13414         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13415         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13416         max skip conditionally.
13417         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13418         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13419         max skip conditionally.
13420         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13421         max skip conditionally.
13422         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13423         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13424         max skip conditionally.
13425         * config/i386/i386.c (struct ptt): Change type from int to
13426         string.
13427         (ix86_default_align): Set default values.
13428         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13429         max skip conditionally.
13430         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13431         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13432         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13433         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13434         max skip conditionally.
13435         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13436         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13437         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13438         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13439         max skip conditionally.
13440         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13441         * config/ia64/ia64.c (ia64_option_override): Set default values
13442         for alignment options.
13443         * config/m68k/m68k.c: Handle new str_align_* options.
13444         * config/mips/mips.c (mips_set_compression_mode): Change
13445         type of constants.
13446         (mips_option_override): Set default values for options.
13447         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13448         Likewise.
13449         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13450         Likewise.
13451         * config/rx/rx.c (rx_option_override): Likewise.
13452         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13453         (LABEL_ALIGN): Use align_labels_log.
13454         (LOOP_ALIGN): Use align_loops_align.
13455         * config/s390/s390.c (s390_asm_output_function_label): Use new
13456         macros.
13457         * config/sh/sh.c (sh_override_options_after_change):
13458         Change type of constants.
13459         * config/spu/spu.c (spu_sched_init): Likewise.
13460         * config/sparc/sparc.c (sparc_option_override): Set default
13461         values for options.
13462         * config/visium/visium.c (visium_option_override): Likewise.
13463         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13464         emit p2align format with last argument if it's not needed.
13465         * doc/invoke.texi: Document extended format of -falign-*.
13466         * final.c: Use align_labels alignment.
13467         * flags.h (struct target_flag_state): Change type to use
13468         align_flags.
13469         (struct align_flags_tuple): New.
13470         (struct align_flags): Likewise.
13471         (align_loops_log): Redefine macro to use new types.
13472         (align_loops_max_skip): Redefine macro to use new types.
13473         (align_jumps_log): Redefine macro to use new types.
13474         (align_jumps_max_skip): Redefine macro to use new types.
13475         (align_labels_log): Redefine macro to use new types.
13476         (align_labels_max_skip): Redefine macro to use new types.
13477         (align_functions_log): Redefine macro to use new types.
13478         (align_loops): Redefine macro to use new types.
13479         (align_jumps): Redefine macro to use new types.
13480         (align_labels): Redefine macro to use new types.
13481         (align_functions): Redefine macro to use new types.
13482         (align_functions_max_skip): Redefine macro to use new types.
13483         (align_loops_value): New macro.
13484         (align_jumps_value): New macro.
13485         (align_labels_value): New macro.
13486         (align_functions_value): New macro.
13487         * function.c (invoke_set_current_function_hook): Propagate
13488         alignment values from flags to global variables default in
13489         topleev.h.
13490         * ipa-icf.c (sem_function::equals_wpa): Use
13491         cl_optimization_option_eq instead of memcmp.
13492         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13493         of string types.
13494         (cl_optimization_stream_in): Likewise.
13495         * optc-save-gen.awk: Support strings in cl_optimization.
13496         * opth-gen.awk: Likewise.
13497         * opts.c (finish_options): Remove error checking of invalid
13498         value ranges.
13499         (MAX_CODE_ALIGN): Remove.
13500         (MAX_CODE_ALIGN_VALUE): Likewise.
13501         (parse_and_check_align_values): New function.
13502         (check_alignment_argument): Likewise.
13503         (common_handle_option): Use check_alignment_argument.
13504         * opts.h (parse_and_check_align_values): Declare.
13505         * toplev.c (init_alignments): Remove.
13506         (read_log_maxskip): New.
13507         (parse_N_M): Likewise.
13508         (parse_alignment_opts): Likewise.
13509         (backend_init_target): Remove usage of init_alignments.
13510         * toplev.h (parse_alignment_opts): Declare.
13511         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13512         argument.
13513         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13514         * tree.c (cl_option_hasher::equal): New.
13515         * varasm.c: Use new global macros.
13517 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13519         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13520         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13521         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13522         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13523         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13524         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13525         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13526         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13527         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13528         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13529         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13530         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13531         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13533 2018-07-04  Martin Liska  <mliska@suse.cz>
13534             Jonathan Wakely  <jwakely@redhat.com>
13536         * coverage.c: Use correct type.
13537         * doc/invoke.texi: Language correction.
13539 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13541         PR target/85620
13542         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13543         ENDBRANCH for non-tail call which may return via indirect branch.
13544         * doc/extend.texi: Document indirect_return attribute.
13546 2018-07-03  Martin Sebor  <msebor@redhat.com>
13548         PR tree-optimization/86274
13549         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13550         precondition.
13551         (format_floating): Correct handling of infinities and NaNs.
13553 2018-07-03  Martin Sebor  <msebor@redhat.com>
13555         * print-tree.c (print_real_cst): New function.
13556         (print_node_brief): Call it.
13557         (print_node): Ditto.
13559 2018-07-03  Jeff Law  <law@redhat.com>
13561         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13562         into a single pattern.
13564         * config/h8300/h8300.md (ors code_iterator): New.
13565         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13566         a single pattern and single splitter.
13567         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13568         (iorqi3_1, xorqi3_1): Likewise.
13569         (iorqi3, xorqi3 expanders): Similarly.
13571         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13572         (movmd_internal) into a single pattern using the P mode iterator.
13573         (movmd splitters): Similarly.
13574         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13575         (movsd splitters): Similarly.
13577         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13578         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13579         right table based on the size of the operand.
13580         * config/h8300/h8300.md (length_table): Corresponding changes. All
13581         references to "addb", "addw" and "addl" changed to "add".
13582         (btst patterns): Merge two variants into a single pattern.
13583         (tstqi, tsthi): Likewise.
13584         (addhi3_incdec, addsi3_incdec): Likewise.
13585         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13586         (mulhi3, mulsi3): Likewise.
13587         (udivhi3, udivsi3): Likewise.
13588         (divhi3, divsi3): Likewise.
13589         (andorqi3, andorhi3, andorsi3): Likewise.
13591 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13593         PR target/85694
13594         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13595         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13597 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13599         PR tree-optimization/85694
13600         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13601         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13602         UNSPEC_URHADD.
13603         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13604         (<u>avg<mode>3_ceil): New patterns.
13606 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13608         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13609         scan-tree-dump directive.
13610         * gcc.dg/vect/slp-perm-2.c: Likewise.
13611         * gcc.dg/vect/slp-perm-3.c: Likewise.
13612         * gcc.dg/vect/slp-perm-5.c: Likewise.
13613         * gcc.dg/vect/slp-perm-6.c: Likewise.
13614         * gcc.dg/vect/slp-perm-7.c: Likewise.
13615         * gcc.dg/vect/slp-perm-8.c: Likewise.
13617 2018-07-03  Marek Polacek  <polacek@redhat.com>
13619         PR middle-end/86202
13620         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13622 2018-07-03  Richard Biener  <rguenther@suse.de>
13624         PR ipa/86389
13625         * tree-ssa-structalias.c (find_func_clobbers): Properly
13626         handle indirect calls.
13628 2018-07-03  Jeff Law  <law@redhat.com>
13630         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13631         (shifts): New code iterator.
13632         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13633         expander.  Fix HImode handling on H8/SX.
13634         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13635         (subqi3, subhi3, subsi3 expanders): Likewise.
13636         (andqi3, andhi3, andsi3 expanders): Likewise.
13637         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13638         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13639         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13640         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13641         (zero_extendqihi2, zero_extendqisi2): Likewise.
13642         (extendqihi2, extendqisi2): Likewise.
13643         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13644         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13645         (rotlqi3_1, rotlhi3_1): Likewise.
13646         (logicalhi3_sn, logicalsi3_sn): Likewise.
13647         (logicalhi3, logicalsi3): Likewise.
13649 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13651         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13652         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13653         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13654         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13655         type to append_pattern_def_seq instead of creating a stmt_vec_info
13656         directly.
13657         (build_mask_conversion): Likewise.  Remove vinfo argument.
13658         (vect_add_conversion_to_patterm): Likewise, renaming to...
13659         (vect_add_conversion_to_pattern): ...this.
13660         (vect_recog_mask_conversion_pattern): Update call to
13661         build_mask_conversion.  Pass the vector type to
13662         append_pattern_def_seq here too.
13663         (vect_recog_gather_scatter_pattern): Update call to
13664         vect_add_conversion_to_pattern.
13666 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13668         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13669         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13670         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13671         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13672         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13673         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13674         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13675         append_pattern_def_seq instead of new_pattern_def_seq.
13676         (vect_recog_divmod_pattern): Do both of the above.
13677         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13678         is null.
13680 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13682         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13683         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13684         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13685         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13686         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13687         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13688         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13689         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13690         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13691         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13692         parameter with a single stmt_vec_info.
13693         (vect_recog_func_ptr): Likewise.
13694         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13695         (vect_try_gather_scatter_pattern): ...this.
13696         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13697         the stmt_vec_info of the statement to be matched.  Don't clear
13698         STMT_VINFO_RELATED_STMT.
13699         (vect_pattern_recog): Update call accordingly.
13701 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13703         PR tree-optimization/85694
13704         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13705         (uavgM3_ceil): Document new optabs.
13706         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13707         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13708         functions.
13709         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13710         (savg_ceil_optab): New optabs.
13711         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13712         (vect_vect_recog_func_ptrs): Add it.
13713         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13714         constant directly from the associated lhs.
13716 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13718         * tree-vect-patterns.c (vect_split_statement): New function.
13719         (vect_convert_input): Use it to try to split an existing cast.
13721 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13723         * poly-int.h (print_hex): New function.
13724         * dumpfile.h (dump_dec, dump_hex): Declare.
13725         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13726         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13727         min_input_precision, operation_precision and operation_sign.
13728         * tree-vect-patterns.c (vect_get_range_info): New function.
13729         (vect_same_loop_or_bb_p, vect_single_imm_use)
13730         (vect_operation_fits_smaller_type): Delete.
13731         (vect_look_through_possible_promotion): Add an optional
13732         single_use_p parameter.
13733         (vect_recog_over_widening_pattern): Rewrite to use new
13734         stmt_vec_info infomration.  Handle one operation at a time.
13735         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13736         (vect_truncatable_operation_p, vect_set_operation_type)
13737         (vect_set_min_input_precision): New functions.
13738         (vect_determine_min_output_precision_1): Likewise.
13739         (vect_determine_min_output_precision): Likewise.
13740         (vect_determine_precisions_from_range): Likewise.
13741         (vect_determine_precisions_from_users): Likewise.
13742         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13743         (vect_vect_recog_func_ptrs): Put over_widening first.
13744         Add cast_forwprop.
13745         (vect_pattern_recog): Call vect_determine_precisions.
13747 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13749         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13750         statements that have been replaced by further pattern statements.
13751         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13753 2018-07-03  Richard Biener  <rguenther@suse.de>
13755         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13756         always set *dt.  Dump vectype in vectype overload.
13757         * dumpfile.h (dump_gimple_expr): New function.
13758         (dump_gimple_expr_loc): Likewise.
13759         * dumpfile.c (dump_gimple_expr): New function.
13760         (dump_gimple_expr_loc): Likewise.
13762 2018-07-02  Jeff Law  <law@redhat.com>
13764         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13765         the H8/300, H8/300H and H8/S variants into a single pattern.
13766         (movhi_h8300, movqi_h8300hs): Similarly.
13767         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13768         (QHI mode iterator): New.
13770         * config/h8300/h8300.md: Remove trailing whitespace.
13772 2018-07-02  Jim Wilson  <jimw@sifive.com>
13774         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13775         instead of emit_insn for interrupt returns.
13776         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13777         (riscv_sret, riscv_uret): Likewise.
13779 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13781         * pretty-print.c (selftest::test_pp_format): Move save and restore
13782         of quotes to class auto_fix_quotes, and add an instance.
13783         * selftest.c: Include "intl.h".
13784         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13785         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13786         * selftest.h (selftest::auto_fix_quotes): New class.
13788 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13790         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13791         (aarch64_sve_prepare_conditional_op): Remove.
13792         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13793         Allow aarch64_simd_reg_or_zero as select operand; remove
13794         the aarch64_sve_prepare_conditional_op call.
13795         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13796         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13797         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13798         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13799         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13800         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13801         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13802         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13803         and a splitters to match all of the *_any patterns.
13804         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13806         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13807         (SVE_COND_FP_BINARY_REV): Remove.
13808         (sve_int_op_rev, sve_fp_op_rev): New.
13809         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13810         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13811         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13812         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13813         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13814         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13815         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13816         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13817         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13819         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13820         Remove match_dup 1 from the inner unspec.
13821         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13823         * config/aarch64/aarch64.md (movprfx): New attr.
13824         (length): Default movprfx to 8.
13825         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13826         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13827         (*<su>mul<SVE_I>3_highpart): Likewise.
13828         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13829         (*v<ASHIFT><SVE_I>3): Likewise.
13830         (*<su><MAXMIN><SVE_I>3): Likewise.
13831         (*<su><MAXMIN><SVE_F>3): Likewise.
13832         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13833         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13834         (*div<SVE_F>4): Likewise.
13836 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13838         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13839         in dump string.
13841 2018-07-02  Richard Biener  <rguenther@suse.de>
13843         PR tree-optimization/86363
13844         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13845         memset argument refers to a non-variable address.
13847 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13849         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13850         {PLUS,MINUS}_EXPR code to...
13851         (adjust_symbolic_bound): ...here,
13852         (combine_bound): ...here,
13853         (set_value_range_with_overflow): ...and here.
13855 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13857         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13858         code...
13859         (extract_range_from_abs_expr): ...here.
13861 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13863         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13864         -fno-omit-frame-pointer when not optimizing.
13866 2018-07-02  Martin Liska  <mliska@suse.cz>
13868         PR ipa/86279
13869         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13870         (propagate_nothrow): Likewise.
13872 2018-07-02  Martin Liska  <mliska@suse.cz>
13874         PR ipa/86323
13875         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13877 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13879         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13880         function in r262149, changing "loc" param from source_location to
13881         const dump_location_t &.
13882         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13883         declaration, as above.
13885 2018-07-01  Paul Koning  <ni1d@arrl.net>
13887         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13888         -munit-asm, -mgnu-asm, -mdec-asm.
13889         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13890         (pdp11_output_labelref): New.
13891         (pdp11_output_def): New.
13892         (pdp11_output_addr_vec_elt): New.
13893         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
13894         %# and %@ format codes.
13895         (pdp11_option_override): New.
13896         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13897         (pdp11_output_ident): New.
13898         (pdp11_asm_named_section): New.
13899         (pdp11_asm_init_sections): New.
13900         (pdp11_file_start): New.
13901         (pdp11_file_end): New.
13902         (output_ascii): Use .ascii/.asciz for -mdec-asm.
13903         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
13904         %o, like %c but octal.
13905         (pdp11_option_override): New.
13906         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13907         -mdec-asm.
13908         (DATA_SECTION_ASM_OP): Ditto.
13909         (READONLY_DATA_SECTION_ASM_OP): New.
13910         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13911         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13912         (ASM_OUTPUT_LABELREF): Ditto.
13913         (ASM_OUTPUT_DEF): Ditto.
13914         (ASM_OUTPUT_EXTERNAL): New.
13915         (ASM_OUTPUT_SOURCE_FILENAME): New.
13916         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13917         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13918         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
13919         %# and %@ format codes.
13920         * config/pdp11/pdp11.opt (mgnu-asm): New.
13921         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13922         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13923         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13925 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
13927         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13928         dereferencing path[] beyond its length.
13929         (debug_path): New.
13930         (debug_all_paths): New.
13931         (rewire_first_differing_edge): New.
13932         (adjust_paths_after_duplication): New.
13933         (duplicate_thread_path): Call adjust_paths_after_duplication.
13934         Add new argument.
13935         (thread_through_all_blocks): Add new argument to
13936         duplicate_thread_path.
13938 2018-06-30  Jim Wilson  <jimw@sifive.com>
13940         * config/riscv/predicates.md (p2m1_shift_operand): New.
13941         (high_mask_shift_operand): New.
13942         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13943         pattern using p2m1_shift_operand.
13944         (lshsi3_zero_extend_3+2): New combiner pattern using
13945         high_mask_shift_operand.
13947 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13949         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13950         split out from...
13951         (vect_recog_rotate_pattern): ...here.
13952         (vect_convert_input): Try to insert casts of invariants in the
13953         preheader.
13954         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13955         preheader to be empty.
13957 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13959         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13960         vector type.  If given, install it in the new statement's
13961         STMT_VINFO_VECTYPE.
13962         (vect_element_precision): New function.
13963         (vect_unpromoted_value): New struct.
13964         (vect_unpromoted_value::vect_unpromoted_value): New function.
13965         (vect_unpromoted_value::set_op): Likewise.
13966         (vect_look_through_possible_promotion): Likewise.
13967         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13968         (vect_widened_op_tree, vect_convert_input): Likewise.
13969         (vect_convert_inputs, vect_convert_output): Likewise.
13970         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13971         to handle the optional cast of the multiplication result and
13972         vect_widened_op_tree to detect the widened multiplication itself.
13973         Do not require the input and output of promotion casts to have
13974         the same sign, but base the signedness of the operation on the
13975         input rather than the result.  If the pattern includes two
13976         promotions, check that those promotions have the same sign.
13977         Do not restrict the MULT_EXPR handling to a double-width result;
13978         handle quadruple-width results and wider.  Use vect_convert_inputs
13979         to convert the inputs to the common type.
13980         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
13981         to handle the optional cast of the ABS result.  Also allow a sign
13982         change or a sign extension between the ABS and MINUS.
13983         Use vect_widened_op_tree to detect the widened subtraction and use
13984         vect_convert_inputs to convert the inputs to the common type.
13985         (vect_handle_widen_op_by_const): Delete.
13986         (vect_recog_widen_op_pattern): New function.
13987         (vect_recog_widen_mult_pattern): Use it.
13988         (vect_recog_widen_shift_pattern): Likewise.
13989         (vect_recog_widen_sum_pattern): Use
13990         vect_look_through_possible_promotion to handle the promoted
13991         PLUS_EXPR operand.
13993 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
13995         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13996         the containing gimple_seq *.
13997         * gimple-iterator.h (gsi_for_stmt): Declare it.
13998         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13999         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14000         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14001         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14002         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14003         checks.
14004         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14005         split out from...
14006         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14007         statement being replaced is part of an existing pattern
14008         definition sequence, inserting the new pattern statements before
14009         the original one.
14010         (vect_pattern_recog_1): Don't return a bool.  If the statement
14011         is already part of a pattern, instead apply pattern matching
14012         to the pattern definition statements.  Don't clear the
14013         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14014         (vect_pattern_recog): Don't break after the first match;
14015         continue processing the pattern definition statements instead.
14016         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14018 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14020         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14021         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14022         (vect_recog_widen_sum_pattern): Use it.
14024 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14026         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14027         phi is not a pattern statement and has not been replaced by
14028         a pattern statement.
14029         * tree-vect-patterns.c (type_conversion_p): Don't check
14030         STMT_VINFO_IN_PATTERN_P.
14031         (vect_recog_vector_vector_shift_pattern): Likewise.
14032         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14033         the pattern statement rather than the original statement; check
14034         directly for a WIDEN_MULT_EXPR here.
14035         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14036         vect_is_simple_use to return the pattern statement rather
14037         than the original statement; use is_pattern_stmt_p to check
14038         for such a pattern statement.
14039         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14040         to return the pattern statement rather than the original statement;
14041         don't do the same transformation here.
14042         (vect_is_simple_use): If the defining statement has been replaced
14043         by a pattern statement, return the pattern statement instead.
14044         Remove the corresponding (local) transformation from the vectype
14045         overload.
14047 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14049         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14050         end and default to null.
14051         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14052         (vectorizable_reduction): Update calls accordingly, dropping the
14053         gimple ** argument if the passed-back statement isn't needed.
14054         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14055         (vect_recog_rotate_pattern): Likewise.
14056         (vect_recog_mask_conversion_pattern): Likewise.
14057         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14058         (vect_mask_constant_operand_p): Likewise.
14059         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14060         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14061         (get_group_load_store_type, get_load_store_type): Likewise.
14062         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14063         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14064         (vectorizable_conversion, vectorizable_assignment): Likewise.
14065         (vectorizable_shift, vectorizable_operation): Likewise.
14066         (vectorizable_store, vect_is_simple_cond): Likewise.
14067         (vectorizable_condition, vectorizable_comparison): Likewise.
14068         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14069         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14070         and move it to the end.  Cope with null def_stmt_outs.
14072 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14074         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14076 2018-06-29  Jeff Law  <law@redhat.com>
14078         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14079         displacements for TARGET_V850E2V3 and newer.
14080         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14081         * config/v850/v850.md (sign23byte_load): Remove.
14082         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14083         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14085 2018-06-29  Martin Liska  <mliska@suse.cz>
14087         PR lto/85759
14088         * coverage.c (coverage_init): Mangle full path name.
14089         * doc/invoke.texi: Document the change.
14090         * gcov-io.c (mangle_path): New.
14091         * gcov-io.h (mangle_path): Likewise.
14092         * gcov.c (mangle_name): Use mangle_path for path mangling.
14094 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14096         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14097         if starting source register is not even.
14099 2018-06-29  Martin Liska  <mliska@suse.cz>
14101         PR tree-optimization/86263
14102         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14103         Make edge redirection.
14105 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14107         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14108         (dump_scope_depth): New variable.
14109         (get_dump_scope_depth): New function.
14110         (dump_begin_scope): New function.
14111         (dump_end_scope): New function.
14112         * dumpfile.h (get_dump_scope_depth): New declaration.
14113         (dump_begin_scope): New declaration.
14114         (dump_end_scope): New declaration.
14115         (class auto_dump_scope): New class.
14116         (AUTO_DUMP_SCOPE): New macro.
14117         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14118         AUTO_DUMP_SCOPE.
14120 2018-06-29  Richard Biener  <rguenther@suse.de>
14122         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14123         compute_all_dependences succeeds.
14124         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14125         exceed --param loop-max-datarefs-for-datadeps.
14127 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14129         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14131 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14133         PR target/86348
14134         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14135         alternative 0 in preferred_for_speed attribute.
14137 2018-06-28  Paul Koning  <ni1d@arrl.net>
14139         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14140         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14141         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14142         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14143         for shift insn patterns.
14145 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14147         * cgraph.c (cgraph_node::get_body): Replace assignments to
14148         "dump_file" with calls to set_dump_file.
14149         * dumpfile.c (alt_dump_file): Make static, and group with...
14150         (alt_flags): ...this definition.
14151         (dumps_are_enabled): New variable.
14152         (refresh_dumps_are_enabled): New function.
14153         (set_dump_file): New function.
14154         (set_alt_dump_file): New function.
14155         (gcc::dump_manager::dump_start): Replace assignments to
14156         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14157         set_alt_dump_file.
14158         (gcc::dump_manager::dump_finish): Likewise.
14159         * dumpfile.h (alt_dump_file): Delete decl.
14160         (dumps_are_enabled): New variable decl.
14161         (set_dump_file): New function decl.
14162         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14163         global.
14164         * tree-nested.c (lower_nested_functions): Replace assignments to
14165         "dump_file" with calls to set_dump_file.
14167 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14169         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14170         goto_locus of each outgoing edge of each basic block.
14172 2018-06-28  Richard Biener  <rguenther@suse.de>
14174         * dwarf2out.c (decl_scope_table): Remove.
14175         (push_decl_scope): Likewise.
14176         (pop_decl_scope): Likewise.
14177         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14178         (gen_struct_or_union_type_die): Likewise.
14179         (gen_tagged_type_die): Likewise.
14180         (dwarf2out_init): Do not initialize decl_scope_table.
14181         (dwarf2out_c_finalize): Do not free it.
14183 2018-06-28  Richard Biener  <rguenther@suse.de>
14185         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14186         deciding whether to not re-use a DIE.
14188 2018-06-28  Richard Biener  <rguenther@suse.de>
14190         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14191         DW_AT_abstract_origin attribute.
14193 2018-06-28  Martin Liska  <mliska@suse.cz>
14195         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14196         Use newly introduced constants.
14197         * tree-switch-conversion.h (struct jump_table_cluster):
14198         Define max_ratio_for_size and max_ratio_for_speed.
14200 2018-06-28  Martin Liska  <mliska@suse.cz>
14202         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14203         Add new checking assert to catch invalid state.
14204         (jump_table_cluster::can_be_handled): Handle single case
14205         clusters.
14206         (jump_table_cluster::is_beneficial): Bail out for such case.
14207         (bit_test_cluster::find_bit_tests):
14208         Add new checking assert to catch invalid state.
14209         (bit_test_cluster::can_be_handled): Handle single case
14210         clusters.
14211         (bit_test_cluster::is_beneficial): Bail out for such case.
14212         (switch_decision_tree::analyze_switch_statement):
14213         Fix comment.
14215 2018-06-28  Martin Liska  <mliska@suse.cz>
14217         * common.opt: Introduce -completion option.
14218         * gcc.c (driver_handle_option): Handle it.
14219         (driver::main): Print completions if completion
14220         is set.
14221         * opt-suggestions.c (option_proposer::get_completions):
14222         New function.
14223         (option_proposer::suggest_completion): Likewise.
14224         (option_proposer::find_param_completions): Likewise.
14225         (verify_autocompletions): Likewise.
14226         (test_completion_valid_options): Likewise.
14227         (test_completion_valid_params): Likewise.
14228         (in_completion_p): Likewise.
14229         (empty_completion_p): Likewise.
14230         (test_completion_partial_match): Likewise.
14231         (test_completion_garbage): Likewise.
14232         (opt_proposer_c_tests): Likewise.
14233         * opt-suggestions.h: Declare new functions.
14234         * opts.c (common_handle_option): Handle OPT__completion_.
14235         * selftest-run-tests.c (selftest::run_tests): Add
14236         opt_proposer_c_tests.
14237         * selftest.c (assert_str_startswith): New.
14238         * selftest.h (assert_str_startswith): Likewise.
14239         (opt_proposer_c_tests): New.
14240         (ASSERT_STR_STARTSWITH): Likewise.
14242 2018-06-28  Martin Liska  <mliska@suse.cz>
14244         * Makefile.in: Add opt-suggestions.o.
14245         * gcc-main.c: Include opt-suggestions.h.
14246         * gcc.c (driver::driver): Likewise.
14247         (driver::~driver): Remove m_option_suggestions.
14248         (driver::build_option_suggestions): Moved to option_proposer.
14249         (driver::suggest_option): Likewise.
14250         (driver::handle_unrecognized_options): Use option_proposer.
14251         * gcc.h (class driver): Add new memver m_option_proposer.
14252         * opt-suggestions.c: New file.
14253         * opt-suggestions.h: New file.
14255 2018-06-28  Martin Liska  <mliska@suse.cz>
14257         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14258         (auto_string_vec::~auto_string_vec): Likewise.
14260 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14262         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14263         prevent_decl_creation_for_types fields up and add reset_location field.
14264         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14265         statement if id->reset_location is true.
14266         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14267         id->reset_location is true.
14268         (copy_phis_for_bb): Force input_location on the arguments if
14269         id->reset_location is true.
14270         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14271         is set on the function to be inlined.
14273 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14275         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14277 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14279         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14280         registers for Pmode.
14281         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14282         hard registers for the clobbered pseudo.
14284 2018-06-27  Paul Koning  <ni1d@arrl.net>
14286         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14287         mutually exclusive options.
14288         * config/pdp11/constraints.md (h): New constraint.
14289         (O): Update definition to match shift code generation.
14290         (D): New constraint.
14291         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14292         (CCFP): Remove.
14293         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14294         function.
14295         (output_jump): Change arguments.
14296         (pdp11_fixed_cc_regs): New function.
14297         (pdp11_cc_mode): Ditto.
14298         (pdp11_expand_shift): Ditto.
14299         (pdp11_assemble_shift): Ditto.
14300         (pdp11_small_shift): Ditto.
14301         (pdp11_branch_cost): Remove.
14302         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14303         from output.
14304         (pdp11_register_move_cost): Update for CC registers.
14305         (pdp11_rtx_costs): Add case for LSHIFTRT.
14306         (pdp11_output_jump): Add CCNZ mode conditional branches.
14307         (notice_update_cc_on_set): Remove.
14308         (pdp11_cc_mode): New function.
14309         (simple_memory_operand): Correct pre/post decrement case.
14310         (no_side_effect_operand): New function.
14311         (pdp11_regno_reg_class): Add CC_REGS class.
14312         (pdp11_fixed_cc_regs): New function.
14313         (pdp11_small_shift): New function.
14314         (pdp11_expand_shift): New function to expand shift insns.
14315         (pdp11_assemble_shift): New function to output shifts.
14316         (pdp11_branch_cost): Remove.
14317         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14318         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14319         (WCHAR_TYPE): Ditto.
14320         (PTRDIFF_TYPE): Ditto.
14321         (ADJUST_INSN_LENGTH): New macro.
14322         (FIXED_REGISTERS): Add CC registers.
14323         (CALL_USED_REGISTERS): Ditto.
14324         (reg_class): Ditto.
14325         (REG_CLASS_NAMES): Ditto.
14326         (REG_CLASS_CONTENTS): Ditto.
14327         (SELECT_CC_MODE): Use new function.
14328         (TARGET_FLAGS_REGNUM): New macro.
14329         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14330         (cc0_reg_rtx): Remove.
14331         (CC_STATUS_MDEP): Remove.
14332         (CC_STATUS_MDEFP_INIT): Remove.
14333         (CC_IN_FPU): Remove.
14334         (NOTICE_UPDATE_CC): Remove.
14335         (REGISTER_NAMES): Add CC registers.
14336         (BRANCH_COST): Change to constant 1.
14337         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14338         handling.
14339         * config/pdp11/pdp11.opt (mbcopy): Remove.
14340         (mbcopy-builtin): Remove.
14341         (mbranch-cheap): Remove.
14342         (mbranch-expensive): Remove.
14343         * config/pdp11/predicates.md (expand_shift_operand): Update to
14344         match shift code generation.
14345         (ccnz_operator): New predicate.
14346         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14347         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14348         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14349         exclusive options.
14350         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14351         description of O constraint.
14353 2018-06-27  Jeff Law  <law@redhat.com>
14354             Austin Law  <austinklaw@gmail.com>
14356         * config/v850/v850.md (addsi3_set_flags): New pattern.
14357         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14358         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14359         (zero_extendhisi2_v850_set_flags): Likewise.
14360         (zero_extendqisi2_v850_set_flags): Likewise.
14361         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14362         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14363         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14365         * config/v850/v850-protos.h (notice_update_cc): Remove.
14366         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14367         (v850_print_operand): Handle 'D' and "d".
14368         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14369         Add handling of arithmetic/logical operations compared against zero.
14370         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14371         Do not look at v850_compare_op, instead get mode from last argument.
14372         (v850_gen_compare): Remove
14373         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14374         after reload for prologue insns.
14375         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14376         patterns.
14377         (construct_save_jarl): Likewise.
14378         (TARGET_FLAGS_REGNUM): Define.
14379         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14380         (NOTICE_UPDATE_CC): Remove.
14381         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14382         than cc0.  Conditionalize on reload_completed.
14383         (cmpsi_insn, setfcc_insn): Likewise.
14384         (tst1 splitter): Turn into define_and_split which sets the flags
14385         after reload.
14386         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14387         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14388         (cstoresf4, cstoredf4): Clobber the flags.
14389         (cmpsi, cmpsf, cmpdf): Remove expanders.
14390         (setf_insn): Remove pattern.
14391         (addsi3): Turn into define_and_split which clobbers the flags after
14392         reload and a suitable pattern (addsi3_clobber_flags) for use after
14393         reload.
14394         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14395         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14396         (ashrsi3, ashrsi3_v850e2): Likewise.
14397         (bins): Clobber the flags.
14398         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14399         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14400         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14401         (call_value_internal_short, call_value_internal_long): Likewise.
14402         (callt_save_interrupt, callt_return_interrupt): Likewise.
14403         (save_interrupt, return_interrupt): Likewise.
14404         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14405         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14406         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14407         (All FP comparisons): Only allow after reload has completed.
14408         (trfsr): Likewise.
14409         (divh, divhu): Tweak output template.
14410         (branch_z_normal, branch_z_invert): Remove
14411         (branch_nz_normal, branch_nz_invert): Likewise.
14412         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14414         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14415         * config/v850/v850.c (notice_update_cc): Remove.
14416         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14417         (CC_NO_CARRY): Likewise.
14418         (NOTICE_UPDATE_CC): Define to nothing.
14419         * config/v850/v850.md: Remove block comment on cc0 handling
14420         Remove "cc" attribute from all patterns.  Remove cc_status handling
14421         from all patterns.  Minor formatting fixes.
14423 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14425         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14426         (cortex-a76.cortex-a55): Likewise.
14427         * config/aarch64/aarch64-tune.md: Regenerate.
14428         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14429         cortex-a76.cortex-a55.
14431 2018-06-27  Jeff Law  <law@redhat.com>
14433         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14434         (MULTILIB_DIRNAMES): Similarly.
14436 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14438         * gimple.h (gimple_return_retbnd): Delete.
14439         (gimple_return_set_retbnd): Likewise.
14440         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14441         gimple_return_set_retbnd.
14442         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14443         gimple_return_retbnd and adjust.
14444         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14445         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14446         Explicitly return NULL in a couple more cases.  Move assertion
14447         on debug statements and remove unreachable code.
14448         (reset_debug_binding): Do not test id->retbnd.
14449         (expand_call_inline): Do not set it.
14451 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14453         * configure.ac: Add --disable-gcov option.
14454         * configure: Regenerate.
14455         * Makefile.in: Honour @enable_gcov@.
14456         * doc/install.texi: Document --disable-gcov.
14458 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14460         * config/arm/arm-cpus.in (cortex-a76): New entry.
14461         (cortex-a76.cortex-a55): Likewise.
14462         * config/arm/arm-tables.opt: Regenerate.
14463         * config/arm/arm-tune.md: Likewise.
14464         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14465         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14466         cortex-a76.cortex-a55.
14468 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14470         PR target/85769
14471         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14473 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14475         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14476         comment.
14477         (EPILOGUE_USES): Likewise.
14479 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14481         * tree-inline.c (remap_location): New function extracted from...
14482         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14483         (copy_phis_for_bb): ...here.  Call remap_location.
14484         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14486 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14488         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14489         unaligned vsx for 16B memset.
14491 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14493         PR target/86285
14494         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14495         ieee128_float_type_node to long_double_type_node unless
14496         TARGET_LONG_DOUBLE_128 is set.
14498 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14500         * cfgloop.c (get_loop_location): Convert return type from
14501         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14502         by implicit construction from rtx_insn *, and using
14503         dump_user_location_t::from_function_decl for the fallback case.
14504         * cfgloop.h (get_loop_location): Convert return type from
14505         location_t to dump_user_location_t.
14506         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14507         dump_printf_loc to pass in a dump_location_t rather than a
14508         location_t, via the gimple stmt.
14509         * coverage.c (get_coverage_counts): Update calls to
14510         dump_printf_loc to pass in dump_location_t rather than a
14511         location_t.
14512         * doc/optinfo.texi (Dump types): Convert example of
14513         dump_printf_loc from taking "locus" to taking "insn".  Update
14514         description of the "_loc" calls to cover dump_location_t.
14515         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14516         "selftest.h".
14517         (dump_user_location_t::dump_user_location_t): New constructors,
14518         from gimple *stmt and rtx_insn *.
14519         (dump_user_location_t::from_function_decl): New function.
14520         (dump_loc): Make static.
14521         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14522         const dump_location_t &.
14523         (dump_generic_expr_loc): Delete.
14524         (dump_printf_loc): Convert param "loc" from location_t to
14525         const dump_location_t &.
14526         (selftest::test_impl_location): New function.
14527         (selftest::dumpfile_c_tests): New function.
14528         * dumpfile.h: Include "profile-count.h".
14529         (class dump_user_location_t): New class.
14530         (struct dump_impl_location_t): New struct.
14531         (class dump_location_t): New class.
14532         (dump_printf_loc): Convert 2nd param from source_location to
14533         const dump_location_t &.
14534         (dump_generic_expr_loc): Delete.
14535         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14536         const dump_location_t &.
14537         * gimple-fold.c (fold_gimple_assign): Update call to
14538         dump_printf_loc to pass in a dump_location_t rather than a
14539         location_t, via the gimple stmt.
14540         (gimple_fold_call): Likewise.
14541         * gimple-loop-interchange.cc
14542         (loop_cand::analyze_iloop_reduction_var): Update for change to
14543         check_reduction_path.
14544         (tree_loop_interchange::interchange): Update for change to
14545         find_loop_location.
14546         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14547         change in return-type of find_loop_location.
14548         (graphite_regenerate_ast_isl): Likewise.
14549         * graphite-optimize-isl.c (optimize_isl): Likewise.
14550         * graphite.c (graphite_transform_loops): Likewise.
14551         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14552         pass in a dump_location_t rather than a location_t, via the
14553         gimple stmt.
14554         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14555         * ipa.c (walk_polymorphic_call_targets): Likewise.
14556         * loop-unroll.c (report_unroll): Convert "locus" param from
14557         location_t to dump_location_t.
14558         (decide_unrolling): Update for change to get_loop_location's
14559         return type.
14560         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14561         location_t to dump_user_location_t.
14562         (grid_find_single_omp_among_assignments_1): Updates calls to
14563         dump_printf_loc to pass in a dump_location_t rather than a
14564         location_t, via the gimple stmt.
14565         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14566         location_t to dump_location_t.  Updates calls to dump_printf_loc
14567         to pass in a dump_location_t rather than a location_t, via the
14568         gimple stmt.
14569         (grid_inner_loop_gridifiable_p): Likewise.
14570         (grid_dist_follows_simple_pattern): Likewise.
14571         (grid_gfor_follows_tiling_pattern): Likewise.
14572         (grid_target_follows_gridifiable_pattern): Likewise.
14573         (grid_attempt_target_gridification): Convert initialization
14574         of local "grid" from memset to zero-initialization; FIXME: does
14575         this require C++11?  Update call to dump_printf_loc to pass in a
14576         optinfo_location rather than a location_t, via the gimple stmt.
14577         * profile.c (read_profile_edge_counts): Updates call to
14578         dump_printf_loc to pass in a dump_location_t rather than a
14579         location_t
14580         (compute_branch_probabilities): Likewise.
14581         * selftest-run-tests.c (selftest::run_tests): Call
14582         dumpfile_c_tests.
14583         * selftest.h (dumpfile_c_tests): New decl.
14584         * tree-loop-distribution.c (pass_loop_distribution::execute):
14585         Update for change in return type of find_loop_location.
14586         * tree-parloops.c (parallelize_loops): Likewise.
14587         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14588         "locus" from location_t to dump_user_location_t.
14589         (canonicalize_loop_induction_variables): Likewise.
14590         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14591         for change in return type of find_loop_location.
14592         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14593         to dump_printf_loc to pass in a dump_location_t rather than a
14594         location_t, via the stmt.
14595         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14596         Likewise.
14597         * tree-vect-loop-manip.c (find_loop_location): Convert return
14598         type from source_location to dump_user_location_t.
14599         (vect_do_peeling): Update for above change.
14600         (vect_loop_versioning): Update for change in type of
14601         vect_location.
14602         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14603         from location_t to dump_user_location_t.
14604         (vect_estimate_min_profitable_iters): Update for change in type
14605         of vect_location.
14606         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14607         location_t to dump_location_t.
14608         (vect_slp_bb): Update for change in type of vect_location.
14609         * tree-vectorizer.c (vect_location): Convert from source_location
14610         to dump_user_location_t.
14611         (try_vectorize_loop_1): Update for change in vect_location's type.
14612         (vectorize_loops): Likewise.
14613         (increase_alignment): Likewise.
14614         * tree-vectorizer.h (vect_location): Convert from source_location
14615         to dump_user_location_t.
14616         (find_loop_location): Convert return type from source_location to
14617         dump_user_location_t.
14618         (check_reduction_path): Convert 1st param from location_t to
14619         dump_user_location_t.
14620         * value-prof.c (check_counter): Update call to dump_printf_loc to
14621         pass in a dump_user_location_t rather than a location_t; update
14622         call to error_at for change in type of "locus".
14623         (check_ic_target): Update call to dump_printf_loc to
14624         pass in a dump_user_location_t rather than a location_t, via the
14625         call_stmt.
14627 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14629         * config/s390/s390.h (enum processor_flags): Do not use
14630         default tune parameter when -march was specified.
14632 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14634         PR target/86314
14635         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14636         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14637         operands.
14639 2018-06-26  Richard Biener  <rguenther@suse.de>
14641         PR tree-optimization/86287
14642         PR bootstrap/86316
14643         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14644         (vect_analyze_loop): Initialize n_stmts.
14646 2018-06-26  Richard Biener  <rguenther@suse.de>
14648         PR middle-end/86271
14649         * fold-const.c (fold_convertible_p): Pointer extension
14650         isn't valid.
14652 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14654         PR debug/86064
14655         * dwarf2out.c (loc_list_has_views): Adjust comments.
14656         (dw_loc_list): Split single cross-partition range with
14657         nonzero locview.
14659 2018-06-25  Jeff Law  <law@redhat.com>
14661         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14662         on -mbig-switch by default.
14664         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14665         test.
14666         (const_float_0_operand): Remove unused predicate.
14667         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14668         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14669         (recipsf2): New expander.  Original pattern now called
14670         (recipsf2_insn).
14671         (recipdf2, recipdf2_insn): Similarly.
14672         (rsqrtsf2, rsqrtsf2_insn): Similarly
14673         (rsqrtdf2, rsqrtdf2_insn): Similarly
14675 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14677         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14678         Simplify logic for FreeBSD (twice).
14680 2018-06-25  Martin Sebor  <msebor@redhat.com>
14682         PR tree-optimization/86204
14683         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14684         a strnlen result if it's less than the length of the string.
14686 2018-06-25  Martin Sebor  <msebor@redhat.com>
14688         PR tree-optimization/85700
14689         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14690         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14691         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14693 2018-06-25  Martin Sebor  <msebor@redhat.com>
14695         * doc/extend.texi (Zero-length arrays): Update and clarify.
14697 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14699         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14700         added IEEE/IBM long double multilib support on PowerPC little
14701         endian Linux systems.
14702         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14703         (MULTILIB_DEFAULTS): Likewise.
14704         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14705         Likewise.
14706         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14707         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14708         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14710 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14712         PR middle-end/86311
14713         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14714         (REORDER_45): Likewise.
14716 2018-06-25  Jeff Law  <law@redhat.com>
14718         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14719         dividend to 32 bits.  Adjust length.
14720         (udivmodhi4): Cleanup output template.  Fix length.
14722 2018-06-25  Carl Love  <cel@us.ibm.com>
14724         * config/rs6000/vsx.md: Change word selector to prefered location.
14726 2018-06-25  Richard Biener  <rguenther@suse.de>
14728         PR tree-optimization/86304
14729         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14730         epilogue-if-converted loops as well.
14732 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14734         * lto-section-out.c (lto_begin_section): Do not print section
14735         name for noaddr and unnumbered dumps.
14737 2018-06-25  Richard Biener  <rguenther@suse.de>
14739         * tree-vectorizer.h (struct vec_info_shared): New structure
14740         with parts split out from struct vec_info and loop_nest from
14741         struct _loop_vec_info.
14742         (struct vec_info): Adjust accordingly.
14743         (struct _loop_vec_info): Likewise.
14744         (LOOP_VINFO_LOOP_NEST): Adjust.
14745         (LOOP_VINFO_DATAREFS): Likewise.
14746         (LOOP_VINFO_DDRS): Likewise.
14747         (struct _bb_vec_info): Likewise.
14748         (BB_VINFO_DATAREFS): Likewise.
14749         (BB_VINFO_DDRS): Likewise.
14750         (struct _stmt_vec_info): Add dr_aux member.
14751         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14752         (DR_MISALIGNMENT_UNINITIALIZED): New.
14753         (set_dr_misalignment): Adjust.
14754         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14755         (vect_analyze_loop): Adjust prototype.
14756         (vect_analyze_loop_form): Likewise.
14757         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14758         Compute dependences lazily.
14759         (vect_record_base_alignments): Use shared datarefs/ddrs.
14760         (vect_verify_datarefs_alignment): Likewise.
14761         (vect_analyze_data_refs_alignment): Likewise.
14762         (vect_analyze_data_ref_accesses): Likewise.
14763         (vect_analyze_data_refs): Likewise.
14764         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14765         constructor parameter for shared part.
14766         (vect_analyze_loop_form): Pass in shared part and adjust.
14767         (vect_analyze_loop_2): Pass in storage for the number of
14768         stmts.  Move loop nest finding to the caller.  Compute
14769         datarefs lazily.
14770         (vect_analyze_loop): Pass in shared part.
14771         (vect_transform_loop): Verify shared datarefs are unchanged.
14772         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14773         constructor parameter for shared part.
14774         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14775         (vect_slp_bb): Verify shared datarefs are unchanged before
14776         transform.
14777         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14778         change.
14779         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14780         DR_MISALIGNMENT_UNINITIALIZED.
14781         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14782         parameter for shared part.
14783         (vec_info::~vec_info): Adjust.
14784         (vec_info_shared::vec_info_shared): New.
14785         (vec_info_shared::~vec_info_shared): Likewise.
14786         (vec_info_shared::save_datarefs): Likewise.
14787         (vec_info_shared::check_datarefs): Likewise.
14788         (try_vectorize_loop_1): Construct shared part live for analyses
14789         of a single loop for multiple vector sizes.
14790         * tree-parloops.c (gather_scalar_reductions): Adjust.
14792 2018-06-25  Richard Biener  <rguenther@suse.de>
14794         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14795         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14796         (vect_analyze_data_refs): Remove similar code from here and
14797         simplify accordingly.
14799 2018-06-25  Richard Biener  <rguenther@suse.de>
14801         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14802         for reverse storage order accesses rather than asserting
14803         they cannot happen here.
14805 2018-06-25  Tom de Vries  <tdevries@suse.de>
14807         PR debug/86257
14808         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14809         Use data16 instead of .byte for insn prefix.
14811 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14813         PR C++/86082
14814         * parser.c (make_char_string_pack): Pass this literal chars
14815         through cpp_interpret_string.
14816         (cp_parser_userdef_numeric_literal): Check the result of
14817         make_char_string_pack.
14819 2018-06-24  Maya Rashish  <coypu@sdf.org>
14821         * ginclude/stddef.h: Simplify conditions around avoiding
14822         re-definition of __size_t.
14824 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14826         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14827         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14829 2018-06-22  Maya Rashish  <coypu@sdf.org>
14831         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14832         made to OpenBSD and NetBSD through the years.
14834 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14836         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14837         behavior of vec_pack (vector double, vector double) to match
14838         behavior of vec_float2 (vector double, vector double).
14840 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14842         * gimplify.c (gimplify_function_tree): Prevent creation
14843         of a trampoline for the address of the current function
14844         passed to entry/exit instrumentation hooks.
14846 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14848         PR target/86222
14849         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14850         correctly.
14852 2018-06-22  Martin Liska  <mliska@suse.cz>
14854         PR tree-optimization/86263
14855         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14856         Bail out if is_enabled is false.
14857         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14858         New declaration.
14859         (jump_table_cluster::is_enabled): New function.
14861 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14863         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14864         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14865         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14866         (lto_input_ts_binfo_tree_pointers): Likewise.
14867         * tree-streamer-out.c (streamer_write_tree_bitfields,
14868         write_ts_binfo_tree_pointers): Likewise.
14869         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14871 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14873         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14875 2018-06-22  Martin Liska  <mliska@suse.cz>
14877         * symbol-summary.h (get): Make it pure and inline move
14878         functionality from ::get function.
14879         (get): Remove and inline into ::get and ::get_create.
14880         (get_create): Move code from ::get function.
14882 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14884         PR target/85994
14885         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14886         -x assembler-with-cpp.
14888 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14890         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14891         _FILE_OFFSET_BITS=64 for C++.
14893 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
14895         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14896         conversion insn that shows up when pr85657-3.c is compiled using
14897         IEEE 128-bit long double.
14898         (neg<mode>2_internal): Use the correct mode to check whether the
14899         mode is IBM extended.
14900         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14901         multiply and divide external functions from being created more
14902         than once.
14904 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14906         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14907         functions.
14908         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
14909         the edge can be forwarded.
14910         (cfg_layout_merge_blocks): Likewise.
14912 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14914         * except.c (finish_eh_generation): Commit edge insertions only after
14915         the EH edges have been redirected from post-landing to landing pads.
14917 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14919         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14920         create_tmp_var_for to create the FRAME decl.
14921         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14923 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
14925         * tree-inline.c (copy_edges_for_bb): Minor tweak.
14926         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14927         debug statement when resetting its value.
14928         (expand_call_inline): Copy the locus of the call onto the assignment
14929         of the return value, if any.  Use local variable in more cases.
14931 2018-06-21  Martin Liska  <mliska@suse.cz>
14933         * ipa-pure-const.c (propagate_nothrow): Use
14934         funct_state_summaries->get.
14935         (dump_malloc_lattice): Likewise.
14936         (propagate_malloc): Likewise.
14938 2018-06-21  Richard Biener  <rguenther@suse.de>
14940         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14941         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14942         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14943         comment.
14944         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14945         BLOCK_ABSTRACT_ORIGIN unconditionally.
14947 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
14949         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14950         deleting it.
14951         * ipa-reference.c (ipa_reference_c_finalize): Delete
14952         ipa_ref_opt_sum_summaries and set it to NULL.
14954 2018-06-21  Tom de Vries  <tdevries@suse.de>
14956         PR tree-optimization/85859
14957         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14958         test with comment from bb_no_side_effects_p.
14960 2018-06-21  Richard Biener  <rguenther@suse.de>
14962         PR tree-optimization/86232
14963         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14964         max for constant niter.
14966 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14968         * config/aarch64/aarch64-simd.md
14969         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14971 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14973         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14974         Make opernads of the unspec commutative.
14976 2018-06-21  Richard Biener  <rguenther@suse.de>
14978         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14979         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14980         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14981         (vect_analyze_data_ref_dependence): Re-order checks to deal with
14982         NULL DR_STEP.
14983         (vect_record_base_alignments): Do not record base alignment
14984         for gathers or scatters.
14985         (vect_compute_data_ref_alignment): Drop return value that is always
14986         true.  Bail out early for gathers or scatters.
14987         (vect_enhance_data_refs_alignment): Bail out early for gathers
14988         or scatters.
14989         (vect_find_same_alignment_drs): Likewise.
14990         (vect_analyze_data_refs_alignment): Remove dead code.
14991         (vect_slp_analyze_and_verify_node_alignment): Likewise.
14992         (vect_analyze_data_refs): For possible gathers or scatters do
14993         not create an alternate DR, just check their possible validity
14994         and mark them.  Adjust DECL_NONALIASED handling to not rely
14995         on DR_BASE_ADDRESS.
14996         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14997         update inits of gathers or scatters.
14998         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14999         Also copy gather/scatter flag to pattern vinfo.
15001 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15003         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15004         behavior of vec_packsu (vector unsigned long long, vector unsigned
15005         long long) to match behavior of vec_packs with same signature.
15007 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15008             Thomas Schwinge <thomas@codesourcery.com>
15009             Cesar Philippidis  <cesar@codesourcery.com>
15011         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15012         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15013         (gimplify_adjust_omp_clauses): Likewise.
15014         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15015         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15016         (gimplify_omp_target_update): Update handling of acc update and
15017         enter/exit data.
15018         * omp-low.c (install_var_field): Remove unused parameter
15019         base_pointers_restrict.
15020         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15021         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15022         FINALIZE}
15023         (omp_target_base_pointers_restrict_p): Delete.
15024         (scan_omp_target): Update call to scan_sharing_clauses.
15025         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15026         FINALIZE}.
15027         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15028         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15029         (convert_local_omp_clauses): Likewise.
15030         * tree-pretty-print.c (dump_omp_clause): Likewise.
15031         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15032         FINALIZE}.
15033         (omp_clause_code_name): Likewise.
15035 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15037         PR debug/86194
15038         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15039         be narrowed.
15041         PR tree-optimization/86231
15042         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15043         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15045 2018-06-20  Tom de Vries  <tdevries@suse.de>
15047         PR tree-optimization/86097
15048         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15049         iv type if signedness of iv type is not the same as that of *nit.
15051 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15053         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15054         EDGE_EH edges, verify they are all EDGE_EH.
15056 2018-06-20  Maya Rashish  <coypu@sdf.org>
15058         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15060 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15062         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15063         * config/aarch64/aarch64.c (xgene1_tunings): Add
15064         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15065         (aarch64_mode_valid_for_sched_fusion_p):
15066         Allow 16-byte modes.
15067         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15068         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15069         128-bit modes.
15070         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15071         New pattern.
15072         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15073         * config/aarch64/iterators.md (VQ2): New mode iterator.
15075 2018-06-20  Martin Liska  <mliska@suse.cz>
15077         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15078         Change default ratio from 10 to 8.
15080 2018-06-20  Martin Liska  <mliska@suse.cz>
15082         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15083         New.
15084         (bit_test_cluster::find_bit_tests): Likewise.
15085         (switch_decision_tree::analyze_switch_statement): Find clusters.
15086         * tree-switch-conversion.h (struct jump_table_cluster): Document
15087         hierarchy.
15089 2018-06-20  Martin Liska  <mliska@suse.cz>
15091         * tree-switch-conversion.c (switch_conversion::collect):
15092         Record m_uniq property.
15093         (switch_conversion::expand): Bail out for special conditions.
15094         (group_cluster::~group_cluster): New.
15095         (group_cluster::group_cluster): Likewise.
15096         (group_cluster::dump): Likewise.
15097         (jump_table_cluster::emit): New.
15098         (switch_decision_tree::fix_phi_operands_for_edges): New.
15099         (struct case_node): Remove struct.
15100         (jump_table_cluster::can_be_handled): New.
15101         (case_values_threshold): Moved to header.
15102         (reset_out_edges_aux): Likewise.
15103         (jump_table_cluster::is_beneficial): New.
15104         (bit_test_cluster::can_be_handled): Likewise.
15105         (add_case_node): Remove.
15106         (bit_test_cluster::is_beneficial): New.
15107         (case_bit_test::cmp): New.
15108         (bit_test_cluster::emit): New.
15109         (expand_switch_as_decision_tree_p): Remove.
15110         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15111         (fix_phi_operands_for_edge): Likewise.
15112         (switch_decision_tree::analyze_switch_statement): New.
15113         (compute_cases_per_edge): Move ...
15114         (switch_decision_tree::compute_cases_per_edge): ... here.
15115         (try_switch_expansion): Likewise.
15116         (switch_decision_tree::try_switch_expansion): Likewise.
15117         (record_phi_operand_mapping): Likewise.
15118         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15119         (emit_case_decision_tree): Likewise.
15120         (switch_decision_tree::emit): Likewise.
15121         (balance_case_nodes): Likewise.
15122         (switch_decision_tree::balance_case_nodes): Likewise.
15123         (dump_case_nodes): Likewise.
15124         (switch_decision_tree::dump_case_nodes): Likewise.
15125         (emit_jump): Likewise.
15126         (switch_decision_tree::emit_jump): Likewise.
15127         (emit_cmp_and_jump_insns): Likewise.
15128         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15129         (emit_case_nodes): Likewise.
15130         (switch_decision_tree::emit_case_nodes): Likewise.
15131         (conditional_probability): Remove.
15132         * tree-switch-conversion.h (enum cluster_type): New.
15133         (PRINT_CASE): New.
15134         (struct cluster): Likewise.
15135         (cluster::cluster): Likewise.
15136         (struct simple_cluster): Likewise.
15137         (simple_cluster::simple_cluster): Likewise.
15138         (struct group_cluster): Likewise.
15139         (struct jump_table_cluster): Likewise.
15140         (struct bit_test_cluster): Likewise.
15141         (struct min_cluster_item): Likewise.
15142         (struct case_tree_node): Likewise.
15143         (case_tree_node::case_tree_node): Likewise.
15144         (jump_table_cluster::case_values_threshold): Likewise.
15145         (struct case_bit_test): Likewise.
15146         (struct switch_decision_tree): Likewise.
15147         (struct switch_conversion): Likewise.
15148         (switch_decision_tree::reset_out_edges_aux): Likewise.
15150 2018-06-20  Martin Liska  <mliska@suse.cz>
15152         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15153         (hoist_edge_and_branch_if_true): Likewise.
15154         (expand_switch_using_bit_tests_p): Likewise.
15155         (struct case_bit_test): Likewise.
15156         (case_bit_test_cmp): Likewise.
15157         (emit_case_bit_tests): Likewise.
15158         (switch_conversion::switch_conversion): New class.
15159         (struct switch_conv_info): Remove old struct.
15160         (collect_switch_conv_info): More to ...
15161         (switch_conversion::collect): ... this.
15162         (check_range): Likewise.
15163         (switch_conversion::check_range): Likewise.
15164         (check_all_empty_except_final): Likewise.
15165         (switch_conversion::check_all_empty_except_final): Likewise.
15166         (check_final_bb): Likewise.
15167         (switch_conversion::check_final_bb): Likewise.
15168         (create_temp_arrays): Likewise.
15169         (switch_conversion::create_temp_arrays): Likewise.
15170         (free_temp_arrays): Likewise.
15171         (gather_default_values): Likewise.
15172         (switch_conversion::gather_default_values): Likewise.
15173         (build_constructors): Likewise.
15174         (switch_conversion::build_constructors): Likewise.
15175         (constructor_contains_same_values_p): Likewise.
15176         (switch_conversion::contains_same_values_p): Likewise.
15177         (array_value_type): Likewise.
15178         (switch_conversion::array_value_type): Likewise.
15179         (build_one_array): Likewise.
15180         (switch_conversion::build_one_array): Likewise.
15181         (build_arrays): Likewise.
15182         (switch_conversion::build_arrays): Likewise.
15183         (gen_def_assigns): Likewise.
15184         (switch_conversion::gen_def_assigns): Likewise.
15185         (prune_bbs): Likewise.
15186         (switch_conversion::prune_bbs): Likewise.
15187         (fix_phi_nodes): Likewise.
15188         (switch_conversion::fix_phi_nodes): Likewise.
15189         (gen_inbound_check): Likewise.
15190         (switch_conversion::gen_inbound_check): Likewise.
15191         (process_switch): Use the newly created class.
15192         (switch_conversion::expand): New.
15193         (switch_conversion::~switch_conversion): New.
15194         * tree-switch-conversion.h: New file.
15196 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15198         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15199         tree-vect-patterns.c.
15200         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15201         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15202         (vect_recog_sad_pattern): Likewise.
15203         (vect_recog_widen_sum_pattern): Likewise.
15204         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15205         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15206         (vect_recog_rotate_pattern): Likewise.
15207         (vect_recog_mult_pattern): Likewise.
15208         (vect_recog_vector_vector_shift_pattern): Likewise.
15209         (vect_recog_divmod_pattern): Likewise.
15210         (vect_recog_mixed_size_cond_pattern): Likewise.
15211         (vect_recog_bool_pattern): Likewise.
15212         (vect_recog_mask_conversion_pattern): Likewise.
15213         (vect_try_gather_scatter_pattern): Likewise.
15214         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15215         (vect_recog_over_widening_pattern): Likewise.
15216         (vect_recog_gather_scatter_pattern): Likewise.
15217         (vect_recog_func_ptr): Move from tree-vectorizer.h
15218         (vect_vect_recog_func_ptrs): Move further down the file.
15219         (vect_recog_func): Likewise.  Remove the third argument.
15220         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15221         (vect_pattern_recog_1): Expect the pattern function to do any
15222         necessary target tests.  Also expect it to provide a vector type.
15223         Remove the type_in handling.
15225 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15227         * tree-vect-patterns.c (vect_pattern_detected): New function.
15228         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15229         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15230         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15231         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15232         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15233         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15234         (vect_recog_mask_conversion_pattern)
15235         (vect_try_gather_scatter_pattern): Likewise.
15237 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15239         * tree-vect-patterns.c (vect_get_internal_def): New function.
15240         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15241         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15242         (search_type_for_mask_1): Use it.
15244 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15246         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15247         redundant WIDEN_SUM_EXPR handling.
15248         (vect_recog_sad_pattern): Likewise.
15250 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15252         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15253         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15254         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15255         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15256         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15258 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15260         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15261         use the stmt_vec_info of the original bb statement for the
15262         new zero assignment, even if the call is part of a pattern.
15264 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15266         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15267         that the sequence is attached to the original statement rather
15268         than the pattern statement.
15269         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15270         PATTERN_DEF_SEQ from the original statement rather than
15271         the main pattern statement.
15272         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15273         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15274         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15276 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15278         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15279         definition statements before the early exit for statements that aren't
15280         live or relevant.
15281         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15282         split out from...
15283         (vect_transform_loop): ...here.  Process pattern definition
15284         statements without first checking whether the main pattern
15285         statement is live or relevant.
15287 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15289         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15290         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15292 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15294         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15295         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15296         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15297         (expand_block_compare): Change select_block_compare_mode call.
15298         (expand_strncmp_align_check): Use new functions, fix comment.
15299         (emit_final_str_compare_gpr): New function.
15300         (expand_strn_compare): Refactor and clean up code.
15301         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15303 2018-06-19  Tony Reix  <tony.reix@atos.com>
15304             Damien Bergamini  <damien.bergamini@atos.com>
15305             David Edelsohn  <dje.gcc@gmail.com>
15307         * collect2.c (static_obj): New variable.
15308         (static_libs): New variable.
15309         (is_in_list): Uncomment declaration.
15310         (main): Track AIX libraries linked statically.
15311         (is_in_list): Uncomment definition.
15312         (scan_prog_file): Don't add AIX shared libraries initializer
15313         to constructor list if linking statically.
15315 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15317         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15318         constant.
15319         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15321 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15323         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15324         blocks.
15326 2018-06-19  Martin Liska  <mliska@suse.cz>
15328         * config/i386/i386.c (ix86_can_inline_p): Do not use
15329         ipa_fn_summaries::get_create.
15330         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15331         get.
15332         (devirtualization_time_bonus): Likewise.
15333         (ipcp_propagate_stage): Likewise.
15334         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15335         (edge_set_predicate): Likewise.
15336         (evaluate_conditions_for_known_args): Likewise.
15337         (evaluate_properties_for_edge): Likewise.
15338         (ipa_call_summary::reset): Tranform to ...
15339         (ipa_call_summary::~ipa_call_summary): ... this.
15340         (ipa_fn_summary::reset): Transform to ...
15341         (ipa_fn_summary::~ipa_fn_summary): ... this.
15342         (ipa_fn_summary_t::remove): Rename to ...
15343         (ipa_fn_summary_t::remove_callees): ... this.
15344         (ipa_fn_summary_t::duplicate): Use placement new
15345         instead of memory copy.
15346         (ipa_call_summary_t::duplicate): Likewise.
15347         (ipa_call_summary_t::remove): Remove.
15348         (dump_ipa_call_summary): Change get_create to get.
15349         (ipa_dump_fn_summary): Dump only when summary exists.
15350         (analyze_function_body): Use symbol_summary::get instead
15351         of get_create.
15352         (compute_fn_summary): Likewise.
15353         (estimate_edge_devirt_benefit): Likewise.
15354         (estimate_edge_size_and_time): Likewise.
15355         (inline_update_callee_summaries): Likewise.
15356         (remap_edge_change_prob): Likewise.
15357         (remap_edge_summaries): Likewise.
15358         (ipa_merge_fn_summary_after_inlining): Likewise.
15359         (write_ipa_call_summary): Likewise.
15360         (ipa_fn_summary_write): Likewise.
15361         (ipa_free_fn_summary): Likewise.
15362         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15363         (struct ipa_call_summary): Likewise.
15364         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15365         of get_create.
15366         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15367         (estimate_size_after_inlining): Likewise.
15368         (estimate_growth): Likewise.
15369         (growth_likely_positive): Likewise.
15370         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15371         (inline_call): Likewise.
15372         * ipa-inline.c (caller_growth_limits): Likewise.
15373         (can_inline_edge_p): Likewise.
15374         (can_inline_edge_by_limits_p): Likewise.
15375         (compute_uninlined_call_time): Likewise.
15376         (compute_inlined_call_time): Likewise.
15377         (want_inline_small_function_p): Likewise.
15378         (edge_badness): Likewise.
15379         (update_caller_keys): Likewise.
15380         (update_callee_keys): Likewise.
15381         (inline_small_functions): Likewise.
15382         (inline_to_all_callers_1): Likewise.
15383         (dump_overall_stats): Likewise.
15384         (early_inline_small_functions): Likewise.
15385         (early_inliner): Likewise.
15386         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15387         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15388         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15389         * ipa-split.c (execute_split_functions): Likewise.
15390         * symbol-summary.h: Likewise.
15391         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15393 2018-06-19  Richard Biener  <rguenther@suse.de>
15395         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15396         (vectorize_loops): ... here.  Fix dbgcnt handling.
15397         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15399 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15401         PR target/86197
15402         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15403         ieee128 argument takes up only one (vector) register, not two (floating
15404         point) registers.
15406 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15408         * gimplify.c (gimplify_init_constructor): Really never clear for an
15409         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15411 2018-06-19  Richard Biener  <rguenther@suse.de>
15413         PR tree-optimization/86179
15414         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15415         after failed recognition.
15417 2018-06-18  Martin Sebor  <msebor@redhat.com>
15419         PR middle-end/85602
15420         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15421         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15422         Handle integer subtraction.
15423         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15424         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15426 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15428         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15429         param from rtx to rtx_insn *.
15430         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15431         param.
15432         (frv_ifcvt_modify_insn): Likwise.
15433         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15434         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15435         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15436         as_a <rtx_insn *> cast to local "unprotected_region" once
15437         it's been established that it's not NULL or pc_rtx.
15438         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15439         param "sethi" from rtx to rtx_insn *.
15440         (nds32_group_float_insns): Likewise for param "insn".
15441         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15442         param.
15443         (vax_output_int_subtract): Likewise.
15444         * config/vax/vax.c (vax_output_int_add): Likewise for param
15445         "insn".
15446         (vax_output_int_subtract): Likewise.
15447         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15448         (emit_pattern_after): Likewise for param "after".
15449         (emit_insn_after): Likewise.
15450         (emit_jump_insn_after): Likewise.
15451         (emit_call_insn_after): Likewise.
15452         (emit_debug_insn_after): Likewise.
15453         (emit_pattern_before): Likewise for param "before".
15454         (emit_insn_before): Likewise.
15455         (emit_jump_insn_before): Likewise.
15456         * final.c (get_insn_template): Likewise for param "insn", removing
15457         a cast.
15458         * output.h (get_insn_template): Likewise for 2nd param.
15459         * rtl.h (emit_insn_before): Likewise.
15460         (emit_jump_insn_before): Likewise.
15461         (emit_debug_insn_before_noloc): Likewise.
15462         (emit_insn_after): Likewise.
15463         (emit_jump_insn_after): Likewise.
15464         (emit_call_insn_after): Likewise.
15465         (emit_debug_insn_after): Likewise.
15466         (set_insn_deleted): Likewise for param.
15468 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15470         PR target/85358
15471         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15472         floating point modes, so that IFmode is numerically greater than
15473         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15474         to declare the ordering.  This prevents IFmode from being
15475         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15476         machine.  Include rs6000-modes.h to share the fractional values
15477         between genmodes* and the rest of the compiler.
15478         (IFmode): Likewise.
15479         (KFmode): Likewise.
15480         (TFmode): Likewise.
15481         * config/rs6000/rs6000-modes.h: New file.
15482         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15483         meaning of rs6000_long_double_size so that 126..128 selects an
15484         appropriate 128-bit floating point type.
15485         (rs6000_option_override_internal): Likewise.
15486         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15487         (TARGET_LONG_DOUBLE_128): Change the meaning of
15488         rs6000_long_double_size so that 126..128 selects an appropriate
15489         128-bit floating point type.
15490         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15491         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15492         source and destination to match the standard usage.
15493         (truncifkf2): Likewise.
15494         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15495         ISA 2.07 to use an explicit clobber, instead of passing in a
15496         temporary.
15497         (copysign<mode>3_soft): Likewise.
15499 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15501         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15502         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15503         (vect_slp_analyze_instance_dependence): Likewise.
15504         (vect_enhance_data_refs_alignment): Likewise.
15505         (vect_analyze_data_refs_alignment): Likewise.
15506         (vect_slp_analyze_and_verify_instance_alignment
15507         (vect_analyze_data_ref_accesses): Likewise.
15508         (vect_prune_runtime_alias_test_list): Likewise.
15509         (vect_analyze_data_refs): Likewise.
15510         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15511         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15512         (vect_analyze_scalar_cycles_1): Likewise.
15513         (vect_get_loop_niters): Likewise.
15514         (vect_analyze_loop_form_1): Likewise.
15515         (vect_update_vf_for_slp): Likewise.
15516         (vect_analyze_loop_operations): Likewise.
15517         (vect_analyze_loop): Likewise.
15518         (vectorizable_induction): Likewise.
15519         (vect_transform_loop): Likewise.
15520         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15521         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15522         (vect_make_slp_decision): Likewise.
15523         (vect_detect_hybrid_slp): Likewise.
15524         (vect_slp_analyze_operations): Likewise.
15525         (vect_slp_bb): Likewise.
15526         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15527         (vectorizable_bswap): Likewise.
15528         (vectorizable_call): Likewise.
15529         (vectorizable_simd_clone_call): Likewise.
15530         (vectorizable_conversion): Likewise.
15531         (vectorizable_assignment): Likewise.
15532         (vectorizable_shift): Likewise.
15533         (vectorizable_operation): Likewise.
15534         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15536 2018-06-18  Martin Sebor  <msebor@redhat.com>
15538         PR tree-optimization/81384
15539         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15540         * builtins.c (expand_builtin_strnlen): New function.
15541         (expand_builtin): Call it.
15542         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15543         * builtins.def (BUILT_IN_STRNLEN): New.
15544         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15545         Warn for bounds in excess of maximum object size.
15546         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15547         single-value ranges.  Handle strnlen.
15548         (handle_builtin_strlen): Handle strnlen.
15549         (strlen_check_and_optimize_stmt): Same.
15550         * doc/extend.texi (Other Builtins): Document strnlen.
15552 2018-06-18  Maya Rashish  <coypu@sdf.org>
15554         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15555         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15556         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15558         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15559         here to ...
15560         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15562 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15564         * tree.c (escaped_string::escape): Replace cast to char * by
15565         const_cast<char *> (unescaped).
15567 2018-06-18  Nick Clifton  <nickc@redhat.com>
15569         PR 84195
15570         * tree.c (escaped_string): New class.  Converts an unescaped
15571         string into its escaped equivalent.
15572         (warn_deprecated_use): Use the new class to convert the
15573         deprecation message, if present.
15574         (test_escaped_strings): New self test.
15575         (test_c_tests): Add test_escaped_strings.
15576         * doc/extend.texi (deprecated): Add a note that the
15577         deprecation message is affected by the -fmessage-length
15578         option, and that control characters will be escaped.
15579         (#pragma GCC error): Document this pragma.
15580         (#pragma GCC warning): Likewise.
15581         * doc/invoke.texi (-fmessage-length): Document this option's
15582         effect on the #warning and #error preprocessor directives and
15583         the deprecated attribute.
15585 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15587         * tree.c (decl_value_expr_lookup): Revert latest change.
15588         (decl_value_expr_insert): Likewise.
15590 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15592         * gimplify.c (nonlocal_vlas): Delete.
15593         (nonlocal_vla_vars): Likewise.
15594         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15595         referenced VLAs.
15596         (gimplify_body): Do not create and destroy nonlocal_vlas.
15597         * tree-nested.c: Include diagnostic.h.
15598         (use_pointer_in_frame): Tweak.
15599         (lookup_field_for_decl): Add assertion and declare the transformation.
15600         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15601         internal error when the reference is in a wrong context.  Do not
15602         create a debug decl by default.
15603         (note_nonlocal_block_vlas): Delete.
15604         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15605         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15606         create a debug decl by default.
15607         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15608         call is in a wrong context.
15609         (fixup_vla_decls): New function.
15610         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15611         debug variables were created.
15612         * tree.c (decl_value_expr_lookup): Add checking assertion.
15613         (decl_value_expr_insert): Likewise.
15615 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15617         PR middle-end/82479
15618         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15619         * tree-scalar-evolution.c (interpret_expr): Likewise.
15620         (expression_expensive_p): Likewise.
15621         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15622         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15623         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15624         (ssa_defined_by_minus_one_stmt_p): New.
15626 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15628         PR middle-end/64946
15629         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15630         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15631         * dojump.c (do_jump): Likewise.
15632         * expr.c (expand_expr_real_2): Check operand type's sign.
15633         * fold-const.c (const_unop): Handle ABSU_EXPR.
15634         (fold_abs_const): Likewise.
15635         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15636         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15637         (strip_sign_op_1): Likesise.
15638         * match.pd: Add new pattern to generate ABSU_EXPR.
15639         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15640         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15641         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15642         * tree-inline.c (estimate_operator_cost): Likewise.
15643         * tree-pretty-print.c (dump_generic_node): Likewise.
15644         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15645         * tree.def (ABSU_EXPR): New.
15647 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15649         PR middle-end/86095
15650         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15651         documented as preserved for backward compatibility only.
15652         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15654         PR rtl-optimization/86108
15655         * bb-reorder.c (create_forwarder_block): Renamed to ...
15656         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15657         jump from new landing pad to the second part.
15658         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15659         Adjust callers.
15661 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15663         PR middle-end/85878
15664         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15665         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15666         Only call store_expr for halves if the mode is the same.
15668         PR middle-end/86123
15669         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15670         Fix up comment formatting.
15672 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15674         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15675         (typed_splay_tree::closure,
15676         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15677         (typed_splay_tree::typed_splay_tree,
15678         typed_splay_tree::operator =): Declared private.
15679         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15680         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15681         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15682         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15683         typed_splay_tree::splay_tree_splay,
15684         typed_splay_tree::splay_tree_foreach_helper,
15685         typed_splay_tree::splay_tree_insert,
15686         typed_splay_tree::splay_tree_remove,
15687         typed_splay_tree::splay_tree_lookup,
15688         typed_splay_tree::splay_tree_predecessor,
15689         typed_splay_tree::splay_tree_successor,
15690         typed_splay_tree::splay_tree_min,
15691         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15692         (typed_splay_tree::root, typed_splay_tree::comp,
15693         typed_splay_tree::delete_key,
15694         typed_splay_tree::delete_value): New data members.
15695         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15696         typed_splay_tree::remove.
15698 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15700         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15701         -mginv and -mno-ginv to the assembler.
15702         * config/mips/mips.opt (-mcrc): New option.
15703         (-mginv): Likewise.
15704         * doc/invoke.text (-mcrc): Document.
15705         (-mginv): Likewise.
15707 2018-06-15  Nick Clifton  <nickc@redhat.com>
15709         PR 84195
15710         * tree.c (escaped_string): New class.  Converts an unescaped
15711         string into its escaped equivalent.
15712         (warn_deprecated_use): Use the new class to convert the
15713         deprecation message, if present.
15714         (test_escaped_strings): New self test.
15715         (test_c_tests): Add test_escaped_strings.
15716         * doc/extend.texi (deprecated): Add a note that the
15717         deprecation message is affected by the -fmessage-length
15718         option, and that control characters will be escaped.
15719         (#pragma GCC error): Document this pragma.
15720         (#pragma GCC warning): Likewise.
15721         * doc/invoke.texi (-fmessage-length): Document this option's
15722         effect on the #warning and #error preprocessor directives and
15723         the deprecated attribute.
15725 2018-06-15  Richard Biener  <rguenther@suse.de>
15727         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15728         here, also noting vector size used.
15729         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15730         size used in MSG_OPTIMIZED_LOCATIONS dump.
15731         (pass_slp_vectorize::execute): Adjust.
15733 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15735         PR target/85968
15736         * config/arc/arc.c (arc_return_address_register): Fix
15737         if-condition.
15739 2018-06-15  Richard Biener  <rguenther@suse.de>
15741         PR middle-end/86159
15742         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15743         leave useless conversion stripping to force_gimple_operand_gsi.
15744         (gimplify_build2): Likewise.
15745         (gimplify_build1): Likewise.
15747 2018-06-15  Richard Biener  <rguenther@suse.de>
15749         PR middle-end/86076
15750         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15751         before adjusting their block.
15753 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15755         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15756         multilibs for *-*-rtems*.
15757         * config/riscv/t-rtems: New file.
15759 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15761         PR middle-end/86122
15762         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15763         unsigned_type_for returns NULL.
15765         PR target/85945
15766         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15767         subregs of multi-word pseudos unless the float mode has word size.
15769 2018-06-14  Richard Biener  <rguenther@suse.de>
15771         PR middle-end/86139
15772         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15773         duplicate and harmful type_hash_canon.
15774         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15776 2018-06-14  Richard Biener  <rguenther@suse.de>
15778         PR ipa/86124
15779         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15780         NULL cgraph_node.
15782 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15784         * config/rtems.h (STDINT_LONG32): Define.
15786 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15787             Prachi Godbole  <prachi.godbole@imgtec.com>
15789         * config/mips/mips-cpus.def: Define P6600.
15790         * config/mips/mips-tables.opt: Regenerate.
15791         * config/mips/mips.c (mips_ucbranch_type): New enum.
15792         (mips_rtx_cost_data): Add support for P6600.
15793         (mips_issue_rate): Likewise.
15794         (mips_multipass_dfa_lookahead): Likewise.
15795         (mips_avoid_hazard): Likewise.
15796         (mips_reorg_process_insns): Likewise.
15797         (mips_classify_branch_p6600): New function.
15798         * config/mips/mips.h (TUNE_P6600): New define.
15799         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15800         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15801         * config/mips/mips.md: Include p6600.md.
15802         (processor): Add p6600.
15803         * config/mips/p6600.md: New file.
15804         * doc/invoke.texi: Add p6600 to supported architectures.
15806 2018-06-13  Martin Sebor  <msebor@redhat.com>
15808         PR tree-optimization/86114
15809         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15810         of integer types.
15811         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15813 2018-06-13  Richard Biener  <rguenther@suse.de>
15815         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15816         Properly set vector type of the intermediate stmt.
15817         * tree-vect-stmts.c (vectorizable_operation): The destination
15818         var always has vectype_out type.
15820 2018-06-13  Jeff Law  <law@redhat.com>
15822         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15823         integer 0 for argument to print_rtl_with_bb.
15824         (rl78_reorg): Likewise.
15826 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15828         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15829         from rtx to rtx_insn *.
15830         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15831         "label".
15832         (add_sched_insns_for_speculation): Likewise for local "target",
15833         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15834         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15835         from rtx_insn ** to rtx_code_label **.
15836         (reorg_emit_nops): Likewise.
15837         (c6x_reorg): Likewise for local "call_labels".
15838         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15839         rtx to rtx_insn *.
15840         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15841         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15842         the loops over LABEL_REFS.
15843         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15844         braf_label.
15845         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15846         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15847         (split_branches): Strengthen local "olabel" from rtx to
15848         rtx_insn *, adding a safe_as_a cast.
15849         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15850         to "rtx_insn *".
15851         (add_insn_after): Likewise for first two params.
15852         (add_insn_before): Likewise.
15853         (remove_insn): Likewise for param.
15854         (emit_pattern_before_noloc): Likewise for second and third params.
15855         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15856         (emit_call_insn_before_noloc): Likewise.
15857         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15858         to "rtx_insn *".
15859         (emit_barrier_before): Likewise.
15860         (emit_label_before): Strengthen "label" param from "rtx" to
15861         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15862         "rtx_insn *".
15863         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15864         "rtx_insn *".
15865         (emit_pattern_after_noloc): Likewise.
15866         (emit_insn_after_noloc): Likewise.
15867         (emit_jump_insn_after_noloc): Likewise.
15868         (emit_call_insn_after_noloc): Likewise.
15869         (emit_debug_insn_after_noloc): Likewise.
15870         (emit_barrier_after): Likewise.
15871         (emit_label_after): Likewise for both params.
15872         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
15873         "loc" param from "int" to "location_t".
15874         (emit_insn_after_setloc): Likewise.
15875         (emit_jump_insn_after_setloc): Likewise.
15876         (emit_call_insn_after_setloc): Likewise.
15877         (emit_debug_insn_after_setloc): Likewise.
15878         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
15879         "loc" param from "int" to "location_t".
15880         (emit_pattern_before): Convert NULL_RTX to NULL.
15881         (emit_insn_before_setloc): Convert "loc" param from "int" to
15882         "location_t".
15883         (emit_jump_insn_before_setloc): Likewise.
15884         (emit_call_insn_before_setloc): Likewise.
15885         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15886         rtx_insn *.  Convert "loc" param from "int" to "location_t".
15887         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15888         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15889         Convert 3rd param from "int" to "location_t".
15890         (emit_barrier_before, emit_barrier_after, next_real_insn):
15891         Strengthen param from rtx to rtx_insn *.
15892         (emit_label_before): Strengthen 1st param from "rtx" to
15893         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
15894         "rtx_insn *".
15895         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15896         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15897         Strengthen 2nd param from "rtx" to "rtx_insn *".
15898         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15899         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15900         Likewise. Convert 3rd param from "int" to "location_t".
15901         (emit_label_after): Strengthen 1st param from "rtx" to
15902         "rtx_code_label *".
15903         (next_real_insn, remove_insn): Strengthen param from "rtx" to
15904         "rtx_insn *".
15905         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15906         from "rtx" to "rtx_insn *".
15908 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
15910         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15911         bodies streamed in with -Q.
15912         * dumpfile.c (dump_files): Add lto-stream-out dump file.
15913         * dumpfile.h (tree_dump_index): Add lto_stream_out.
15914         * gimple-streamer-out.c: Include gimple-pretty-print.h
15915         (output_bb): Dump stmts streamed.
15916         * lto-section-out.c: Include print-tree.h
15917         (lto_begin_section): Dump sections created.
15918         (lto_output_decl_index): Dump decl encoded.
15919         * lto-streamer-out.c: Include print-tree.h
15920         (create_output_block): Dump output block created.
15921         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15922         (output_function): Dump function output.
15923         (output_constructor): Dump constructor streamed.
15924         (write_global_stream): Output indexes encoded.
15925         (produce_asm_for_decls): Dump streams encoded.
15926         * lto-streamer.c (streamer_dump_file): New global var.
15927         * lto-streamer.h (streamer_dump_file): Declare.
15928         * passes.c (ipa_write_summaries): Initialize streamer dump.
15929         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15930         in.
15932 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
15934         PR target/86048
15935         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15936         offsets for register save directives.  Emit a second batch of save
15937         directives, if need be, when the function accesses prior frames.
15939 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15941         * config/arc/fpu.md (fmasf4): Force operand to register.
15942         (fnmasf4): Likewise.
15944 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15946         * config/arc/arc-protos.h (arc_pad_return): Remove.
15947         * config/arc/arc.c (machine_function): Remove force_short_suffix
15948         and size_reason.
15949         (arc_print_operand): Adjust printing of '&'.
15950         (arc_verify_short): Remove conditional printing of short suffix.
15951         (arc_final_prescan_insn): Remove reference to size_reason.
15952         (pad_return): New function.
15953         (arc_reorg): Call pad_return.
15954         (arc_pad_return): Remove.
15955         (arc_init_machine_status): Remove reference to force_short_suffix.
15956         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15957         (attr length): When attribute iscompact is true force to 2
15958         regardless; in the case of maybe check if we want to force the
15959         instruction to have 4 bytes length.
15960         (nopv): Change it to generate 4 byte long nop as well.
15961         (blockage): New pattern.
15962         (simple_return): Remove call to arc_pad_return.
15963         (p_return_i): Likewise.
15965 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15967         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15969 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15971         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15972         ARC cores.
15974 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15976         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15977         for ARC700 and ARCv2.
15979 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
15981         PR target/86076
15982         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15983         operands[2] instead of operands[1].
15986 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15988         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15989         case, check whether the outer register overlaps an unallocatable
15990         register, not just whether it fits the required class.
15992 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
15994         * poly-int.h (can_div_trunc_p): Add new overload in which all values
15995         are poly_ints.
15996         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15997         (memrefs_conflict_p): Likewise.
15998         (init_alias_analysis): Likewise.
15999         * cfgexpand.c (expand_debug_expr): Likewise.
16000         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16001         * cse.c (fold_rtx): Likewise.
16002         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16003         * expr.c (emit_block_move_hints): Likewise.
16004         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16005         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16006         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16007         (emit_group_store): Likewise.
16008         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16009         to read the PRE/POST_MODIFY increment.
16010         * calls.c (store_one_arg): Use strip_offset.
16011         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16012         poly_int_rtx_p.
16013         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16014         by a VEC_SELECT.
16015         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16016         (simplify_binary_operation_1): Extend CONST_INT handling to
16017         poly_int_rtx_p.
16018         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16019         than a HOST_WIDE_INT.
16020         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16021         poly_int64.
16022         (adjust_mems, add_stores): Update accodingly.
16023         (vt_canonicalize_addr): Track polynomial offsets.
16024         (emit_note_insn_var_location): Likewise.
16025         (vt_add_function_parameter): Likewise.
16026         (vt_initialize): Likewise.
16028 2018-06-12  Jeff Law  <law@redhat.com>
16030         * config.gcc (alpha*-*-freebsd*): Remove.
16031         * config/alpha/freebsd.h: Remove.
16033 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16035         PR other/69968
16036         * spellcheck-tree.c (levenshtein_distance): Rename to...
16037         (get_edit_distance): ...this, and update for underlying renaming.
16038         * spellcheck-tree.h (levenshtein_distance): Rename to...
16039         (get_edit_distance): ...this.
16040         * spellcheck.c (levenshtein_distance): Rename to...
16041         (get_edit_distance): ...this.  Convert from Levenshtein distance
16042         to Damerau-Levenshtein distance by supporting transpositions of
16043         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16044         "v_one_ago".
16045         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16046         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16047         update for underlying renaming.
16048         (selftest::levenshtein_distance_unit_test): Rename to...
16049         (selftest::test_get_edit_distance_unit): ...this, and update for
16050         underlying renaming.
16051         (selftest::test_find_closest_string): Add example from PR 69968
16052         where transposition helps
16053         (selftest::test_metric_conditions): Update for renaming.
16054         (selftest::test_metric_conditions): Likewise.
16055         (selftest::spellcheck_c_tests): Likewise.
16056         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16057         (get_edit_distance): ...this.
16058         (best_match::consider): Update for renaming.
16060 2018-06-12  Martin Sebor  <msebor@redhat.com>
16062         PR tree-optimization/85259
16063         * builtins.c (compute_objsize): Handle constant offsets.
16064         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16065         true iff a warning has been issued.
16066         * gimple.h (gimple_nonartificial_location): New function.
16067         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16068         gimple_nonartificial_location and handle -Wno-system-headers.
16069         (handle_builtin_stxncpy): Same.
16071 2018-06-12  Martin Sebor  <msebor@redhat.com>
16073         PR c/85931
16074         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16076 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16078         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16079         BUILTIN_VEC_XST entries for pointer to double and long long.
16081 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16083         PR target/85990
16084         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16085         Update comments.
16086         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16087         Likewise.
16089 2018-06-12  Martin Liska  <mliska@suse.cz>
16091         * doc/options.texi: Document IntegerRange.
16093 2018-06-12  Martin Liska  <mliska@suse.cz>
16095         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16096         * opt-functions.awk: Handle Deprecated flag.
16097         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16098         and report error.
16099         (read_cmdline_option): Report warning for a deprecated option.
16100         * opts.h (struct cl_option): Add new field cl_deprecated.
16101         (CL_ERR_DEPRECATED): New.
16103 2018-06-12  Martin Liska  <mliska@suse.cz>
16105         * doc/options.texi: Document Deprecated option flag.
16107 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16109         * config/arc/arc-arch.h (arc_extras): New enum.
16110         (arc_cpu_t):Add field extra.
16111         (arc_cpu_types): Consider the extras.
16112         * config/arc/arc-cpus.def: Add extras info.
16113         * config/arc/arc-opts.h (processor_type): Consider extra field.
16114         * config/arc/arc.c (arc_override_options): Handle extra field.
16116 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16118         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16119         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16120         field.
16121         * config/arc/arc.c (arc_init): Update pic warning.
16122         (irq_range): Update irq range parsing warnings.
16123         (arc_override_options): Update various warning messages.
16124         (arc_handle_aux_attribute): Likewise.
16126 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16128         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16130 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16132         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16133         compiler messages directives.
16135 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16137         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16138         * config/mips/mips-tables.opt: Regenerate.
16139         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16140         mips64r6.
16141         * doc/invoke.texi: Document -march=i6500.
16143 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16145         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16146         (i6400_gpmul): Add cpu_unit.
16147         (i6400_gpdiv): Likewise.
16148         (i6400_msa_add_d): Update reservations.
16149         (i6400_msa_int_add) Likewise.
16150         (i6400_msa_short_logic3) Likewise.
16151         (i6400_msa_short_logic2) Likewise.
16152         (i6400_msa_short_logic) Likewise.
16153         (i6400_msa_move) Likewise.
16154         (i6400_msa_cmp) Likewise.
16155         (i6400_msa_short_float2) Likewise.
16156         (i6400_msa_div_d) Likewise.
16157         (i6400_msa_long_logic1) Likewise.
16158         (i6400_msa_long_logic2) Likewise.
16159         (i6400_msa_mult) Likewise.
16160         (i6400_msa_long_float2) Likewise.
16161         (i6400_msa_long_float4) Likewise.
16162         (i6400_msa_long_float5) Likewise.
16163         (i6400_msa_long_float8) Likewise.
16164         (i6400_fpu_fadd): Include frint type.
16165         (i6400_fpu_store): New define_insn_reservation.
16166         (i6400_fpu_load): Likewise.
16167         (i6400_fpu_move): Likewise.
16168         (i6400_fpu_fcmp): Likewise.
16169         (i6400_fpu_fmadd): Likewise.
16170         (i6400_int_mult): Include imul3nc type and update reservation.
16171         (i6400_int_div): Include idiv3 type and update reservation.
16172         (i6400_int_load): Update to check type not move_type.
16173         (i6400_int_store): Likewise.
16174         (i6400_int_prefetch): Set zero latency.
16176 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16178         * gcc.c: Document new %@{...} sequence.
16179         (LINK_COMMAND_SPEC): Use it for the -L switches.
16180         (cpp_unique_options): Use it for the -I switches.
16181         (at_file_argbuf): New global variable.
16182         (in_at_file): Likewise.
16183         (alloc_args): Create at_file_argbuf.
16184         (clear_args): Truncate at_file_argbuf.
16185         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16186         (open_at_file): New function.
16187         (close_at_file): Likewise.
16188         (create_at_file): Delete.
16189         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16190         <'o'>: Likewise.
16191         <'@'>: New case.
16192         (validate_switches_from_spec): Deal with %@{...} sequence.
16193         (validate_switches): Likewise.
16194         (driver::finalize): Call clear_args.
16196 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16198         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16200 2018-06-11  Martin Sebor  <msebor@redhat.com>
16202         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16203         (Wno-class-memaccess): Add @opindex.
16204         (Wno-templates, Wno-multiple-inheritance): Same.
16205         (Wno-virtual-inheritance, Wno-namespaces): Same.
16206         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16207         (Wno-format-overflow, Wno-format-truncation): Same.
16208         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16209         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16210         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16211         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16212         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16213         (Wno-misspelled-isr): Same.
16215 2018-06-11  Martin Sebor  <msebor@redhat.com>
16217         * PR tree-optimization/86083
16218         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16220 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16222         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16224 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16226         PR target/85755
16227         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16228         on the correct operand.
16229         (*movdi_internal64): Ditto.
16231 2018-06-11  Martin Liska  <mliska@suse.cz>
16233         PR tree-optimization/86089
16234         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16236 2018-06-11  Julia Koval  <julia.koval@intel.com>
16238         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16239         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16240         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16242 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16244         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16245         for Ada with strict dwarf2.
16247 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16249         PR target/85755
16250         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16251         addresses.
16253 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16255         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16257 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16259         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16260         TARGET_ELF.
16262 2018-06-08  Martin Liska  <mliska@suse.cz>
16264         * tree-cfg.h (debug_function): Fix argument type to match
16265         implementation.
16267 2018-06-08  Martin Liska  <mliska@suse.cz>
16269         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16270         Remove usage of MPX-related (and removed) fields.
16271         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16273 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16275         * cfg.c (debug): Use TDF_NONE rather than 0.
16276         * cfghooks.c (debug): Likewise.
16277         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16278         (struct dump_option_value_info): Convert to...
16279         (struct kv_pair): ...this template type.
16280         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16281         rather than 0.
16282         (optinfo_verbosity_options): Likewise.
16283         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16284         OPTGROUP_NONE.
16285         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16286         than int for "optgroup_flags" param.
16287         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16288         "dump_kind" param.
16289         (dump_dec): Likewise.
16290         (dump_finish): Use TDF_NONE rather than 0.
16291         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16292         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16293         than 0.  Update for change to option_ptr.
16294         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16295         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16296         0.  Update for changes to optinfo_verbosity_options and
16297         optgroup_options.
16298         (opt_info_switch_p): Convert optgroup_flags from int to
16299         optgroup_flags_t.
16300         (dump_basic_block): Use dump_flags_t rather than int
16301         for "dump_kind" param.
16302         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16303         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16304         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16305         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16306         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16307         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16308         TDF_NONE): Convert from macros to...
16309         (enum dump_flag): ...this new enum.
16310         (dump_flags_t): Update to use enum.
16311         (operator|, operator&, operator~, operator|=, operator&=):
16312         Implement for dump_flags_t.
16313         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16314         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16315         Convert from macros to...
16316         (enum optgroup_flag): ...this new enum.
16317         (optgroup_flags_t): New typedef.
16318         (operator|, operator|=): Implement for optgroup_flags_t.
16319         (struct dump_file_info): Convert field "alt_flags" to
16320         dump_flags_t.  Convert field "optgroup_flags" to
16321         optgroup_flags_t.
16322         (dump_basic_block): Use dump_flags_t rather than int for param.
16323         (dump_generic_expr_loc): Likewise.
16324         (dump_dec): Likewise.
16325         (dump_register): Convert param "optgroup_flags" to
16326         optgroup_flags_t.
16327         (opt_info_enable_passes): Likewise.
16328         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16329         than 0.
16330         * gimple-pretty-print.c (debug): Likewise.
16331         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16332         (merged_store_group::apply_stores): Likewise.
16333         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16334         * gimple.c (verify_gimple_pp): Likewise.
16335         * graphite-poly.c (print_pbb_body): Likewise.
16336         * passes.c (pass_manager::register_one_dump_file): Convert
16337         local "optgroup_flags" to optgroup_flags_t.
16338         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16339         (debug): Likewise.
16340         (debug_body): Likewise.
16341         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16342         to optgroup_flags_t.
16343         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16344         than 0.
16345         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16346         (convert_mult_to_fma): Likewise.
16347         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16348         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16349         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16350         "dump_kind" to dump_flags_t.
16352 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16354         * config/rs6000/rs6000.c (min, max): Delete.
16356 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16358         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16359         -mabi=spe and -mabi=no-spe.
16361 2018-06-08  Martin Liska  <mliska@suse.cz>
16363         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16364         where we expect an existing summary.
16366 2018-06-08  Martin Liska  <mliska@suse.cz>
16368         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16369         * ipa-inline.h (estimate_edge_growth): Likewise.
16371 2018-06-08  Martin Liska  <mliska@suse.cz>
16373         * cgraph.c (function_version_hasher::hash): Use
16374         cgraph_node::get_uid ().
16375         (function_version_hasher::equal):
16376         * cgraph.h (cgraph_node::get_uid): New method.
16377         * ipa-inline.c (update_caller_keys): Use
16378         cgraph_node::get_uid ().
16379         (update_callee_keys): Likewise.
16380         * ipa-utils.c (searchc): Likewise.
16381         (ipa_reduced_postorder): Likewise.
16382         * lto-cgraph.c (input_node): Likewise.
16383         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16384         * symbol-summary.h (symtab_insertion): Likewise.
16385         (symtab_removal): Likewise.
16386         (symtab_duplication): Likewise.
16387         * tree-pretty-print.c (dump_function_header): Likewise.
16388         * tree-sra.c (convert_callers_for_node): Likewise.
16390 2018-06-08  Martin Liska  <mliska@suse.cz>
16392         * cgraph.c (symbol_table::create_edge): Always assign a new
16393         unique number.
16394         (symbol_table::free_edge): Do not recycle numbers.
16395         * cgraph.h (cgraph_edge::get): New method.
16396         * symbol-summary.h (symtab_removal): Use it.
16397         (symtab_duplication): Likewise.
16398         (call_summary::hashable_uid): Remove.
16400 2018-06-08  Martin Liska  <mliska@suse.cz>
16402         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16403         (initialize_growth_caches): Remove.
16404         (free_growth_caches): Likewise.
16405         (do_estimate_edge_time): Use edge_growth_cache.
16406         (do_estimate_edge_size): Likewise.
16407         (do_estimate_edge_hints): Likewise.
16408         * ipa-inline.c (reset_edge_caches): Likewise.
16409         (recursive_inlining): Likewise.
16410         (inline_small_functions): Likewise.
16411         * ipa-inline.h (initialize_growth_caches): Remove.
16412         (estimate_edge_size): Likewise.
16413         (estimate_edge_time): Likewise.
16414         (estimate_edge_hints): Likewise.
16415         (reset_edge_growth_cache): Likewise.
16416         * symbol-summary.h (call_summary::remove): New method.
16418 2018-06-08  Martin Liska  <mliska@suse.cz>
16420         * ipa-cp.c (class edge_clone_summary): New summary.
16421         (grow_edge_clone_vectors): Remove.
16422         (ipcp_edge_duplication_hook): Remove.
16423         (class edge_clone_summary_t): New call_summary class.
16424         (ipcp_edge_removal_hook): Remove.
16425         (edge_clone_summary_t::duplicate): New function.
16426         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16427         (create_specialized_node): Likewise.
16428         (ipcp_driver): Initialize edge_clone_summaries and do not
16429         register hooks.
16431 2018-06-08  Martin Liska  <mliska@suse.cz>
16433         * symbol-summary.h (get): New function.
16434         (call_summary::m_initialize_when_cloning): New class member.
16436 2018-06-08  Martin Liska  <mliska@suse.cz>
16438         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16439         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16440         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16441         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16442         (remove_cgraph_node_from_order): Use the removed_nodes set.
16443         (do_per_function_toporder): Likwise.
16444         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16445         instead of summary_uid.
16446         (symtab_removal): Likewise.
16447         (symtab_duplication): Likewise.
16449 2018-06-08  Martin Liska  <mliska@suse.cz>
16451         * ipa-cp.c (ipcp_store_bits_results): Use
16452         ipcp_transformation_sum.
16453         (ipcp_store_vr_results): Likewise.
16454         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16455         to ...
16456         (ipcp_transformation_initialize): ... this.
16457         (ipa_set_node_agg_value_chain):
16458         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16459         (write_ipcp_transformation_info): Likewise.
16460         (read_ipcp_transformation_info): Likewise.
16461         (ipcp_update_bits): Likewise.
16462         (ipcp_update_vr): Likewise.
16463         (ipcp_transform_function): Likewise.
16464         * ipa-prop.h: Rename ipcp_transformation_summary to
16465         ipcp_transformation.
16466         (class ipcp_transformation_t): New function summary.
16467         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16468         (ipa_get_agg_replacements_for_node): Likewise.
16470 2018-06-08  Martin Liska  <mliska@suse.cz>
16472         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16473         of struct.
16474         (class funct_state_summary_t): New function_summary class.
16475         (has_function_state): Remove.
16476         (get_function_state): Likewise.
16477         (set_function_state): Likewise.
16478         (add_new_function): Likewise.
16479         (funct_state_summary_t::insert): New function.
16480         (duplicate_node_data): Remove.
16481         (remove_node_data): Remove.
16482         (funct_state_summary_t::duplicate): New function.
16483         (register_hooks): Create new funct_state_summaries.
16484         (pure_const_generate_summary): Use it.
16485         (pure_const_write_summary): Likewise.
16486         (pure_const_read_summary): Likewise.
16487         (propagate_pure_const): Likewise.
16488         (propagate_nothrow): Likewise.
16489         (dump_malloc_lattice): Likewise.
16490         (propagate_malloc): Likewise.
16491         (execute): Do not register hooks, just remove summary
16492         instead.
16493         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16494         constructor.
16496 2018-06-08  Martin Liska  <mliska@suse.cz>
16498         * ipa-reference.c (remove_node_data): Remove.
16499         (duplicate_node_data): Likewise.
16500         (class ipa_ref_var_info_summary_t): New class.
16501         (class ipa_ref_opt_summary_t): Likewise.
16502         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16503         (get_reference_optimization_summary): Use
16504         ipa_ref_opt_sum_summaries.
16505         (set_reference_vars_info): Remove.
16506         (set_reference_optimization_summary): Likewise.
16507         (ipa_init): Create summaries.
16508         (init_function_info): Use function summary.
16509         (ipa_ref_opt_summary_t::duplicate): New function.
16510         (ipa_ref_opt_summary_t::remove): New function.
16511         (get_read_write_all_from_node): Fix GNU coding style.
16512         (propagate): Use function summary.
16513         (write_node_summary_p): Fix GNU coding style.
16514         (stream_out_bitmap): Likewise.
16515         (ipa_reference_read_optimization_summary): Use function summary.
16516         (ipa_reference_c_finalize): Do not release hooks.
16518 2018-06-08  Martin Liska  <mliska@suse.cz>
16520         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16521         (analyze_function_body): Extract multiple calls of get_create.
16522         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16523         * ipa-inline.c (recursive_inlining): Use ::get method.
16524         * ipa-inline.h (estimate_edge_growth): Likewise.
16526 2018-06-08  Martin Liska  <mliska@suse.cz>
16528         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16529         HSA_INVALID.
16530         (hsa_function_summary::hsa_function_summary): Use the new enum
16531         value.
16532         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16533         * hsa-gen.c (hsa_get_host_function): Likewise.
16534         (get_brig_function_name): Likewise.
16535         * ipa-hsa.c (process_hsa_functions): Likewise.
16536         (ipa_hsa_write_summary): Likewise.
16537         * symbol-summary.h (symtab_duplication): Use ::get function/
16538         (get): New function.
16540 2018-06-08  Martin Liska  <mliska@suse.cz>
16542         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16543         of get.
16544         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16545         (hsa_register_kernel): Likewise.
16546         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16547         * hsa-gen.c (hsa_get_host_function): Likewise.
16548         (get_brig_function_name): Likewise.
16549         (generate_hsa): Likewise.
16550         (pass_gen_hsail::execute): Likewise.
16551         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16552         (devirtualization_time_bonus): Likewise.
16553         (ipcp_propagate_stage): Likewise.
16554         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16555         (edge_set_predicate): Likewise.
16556         (evaluate_conditions_for_known_args): Likewise.
16557         (evaluate_properties_for_edge): Likewise.
16558         (ipa_fn_summary::reset): Likewise.
16559         (ipa_fn_summary_t::duplicate): Likewise.
16560         (dump_ipa_call_summary): Likewise.
16561         (ipa_dump_fn_summary): Likewise.
16562         (analyze_function_body): Likewise.
16563         (compute_fn_summary): Likewise.
16564         (estimate_edge_devirt_benefit): Likewise.
16565         (estimate_edge_size_and_time): Likewise.
16566         (estimate_calls_size_and_time): Likewise.
16567         (estimate_node_size_and_time): Likewise.
16568         (inline_update_callee_summaries): Likewise.
16569         (remap_edge_change_prob): Likewise.
16570         (remap_edge_summaries): Likewise.
16571         (ipa_merge_fn_summary_after_inlining): Likewise.
16572         (ipa_update_overall_fn_summary): Likewise.
16573         (read_ipa_call_summary): Likewise.
16574         (inline_read_section): Likewise.
16575         (write_ipa_call_summary): Likewise.
16576         (ipa_fn_summary_write): Likewise.
16577         (ipa_free_fn_summary): Likewise.
16578         * ipa-hsa.c (process_hsa_functions): Likewise.
16579         (ipa_hsa_write_summary): Likewise.
16580         (ipa_hsa_read_section): Likewise.
16581         * ipa-icf.c (sem_function::merge): Likewise.
16582         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16583         (do_estimate_edge_time): Likewise.
16584         (estimate_size_after_inlining): Likewise.
16585         (estimate_growth): Likewise.
16586         (growth_likely_positive): Likewise.
16587         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16588         (inline_call): Likewise.
16589         * ipa-inline.c (caller_growth_limits): Likewise.
16590         (can_inline_edge_p): Likewise.
16591         (can_inline_edge_by_limits_p): Likewise.
16592         (compute_uninlined_call_time): Likewise.
16593         (compute_inlined_call_time): Likewise.
16594         (want_inline_small_function_p): Likewise.
16595         (edge_badness): Likewise.
16596         (update_caller_keys): Likewise.
16597         (update_callee_keys): Likewise.
16598         (recursive_inlining): Likewise.
16599         (inline_small_functions): Likewise.
16600         (inline_to_all_callers_1): Likewise.
16601         (dump_overall_stats): Likewise.
16602         (early_inline_small_functions): Likewise.
16603         (early_inliner): Likewise.
16604         * ipa-inline.h (estimate_edge_growth): Likewise.
16605         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16606         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16607         * ipa-prop.h (IPA_NODE_REF): Likewise.
16608         (IPA_EDGE_REF): Likewise.
16609         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16610         (propagate_malloc): Likewise.
16611         * ipa-split.c (execute_split_functions): Likewise.
16612         * symbol-summary.h: Rename get to get_create.
16613         (get): Likewise.
16614         (get_create): Likewise.
16615         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16617 2018-06-08  Martin Liska  <mliska@suse.cz>
16619         * symbol-summary.h (release): Move definition out of class
16620         declaration.
16621         (symtab_removal): Likewise.
16622         (symtab_duplication): Likewise.
16624 2018-06-08  Martin Liska  <mliska@suse.cz>
16626         * symbol-summary.h (function_summary): Move constructor
16627         implementation out of class declaration.
16628         (release): Likewise.
16629         (symtab_insertion): Likewise.
16630         (symtab_removal): Likewise.
16631         (symtab_duplication): Likewise.
16632         (get): Likewise.
16634 2018-06-08  Martin Liska  <mliska@suse.cz>
16636         * Makefile.in: Remove support for MPX (macros, related functions,
16637         fields in cgraph_node, ...).
16638         * builtin-types.def (BT_BND): Likewise.
16639         (BT_FN_BND_CONST_PTR): Likewise.
16640         (BT_FN_CONST_PTR_BND): Likewise.
16641         (BT_FN_VOID_PTR_BND): Likewise.
16642         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16643         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16644         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16645         (expand_builtin_mempcpy_with_bounds): Likewise.
16646         (expand_builtin_memset_with_bounds): Likewise.
16647         (expand_builtin_memset_args): Likewise.
16648         (std_expand_builtin_va_start): Likewise.
16649         (expand_builtin): Likewise.
16650         (expand_builtin_with_bounds): Likewise.
16651         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16652         (DEF_LIB_BUILTIN_CHKP): Likewise.
16653         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16654         (DEF_CHKP_BUILTIN): Likewise.
16655         (BUILT_IN_MEMCPY): Likewise.
16656         (BUILT_IN_MEMMOVE): Likewise.
16657         (BUILT_IN_MEMPCPY): Likewise.
16658         (BUILT_IN_MEMSET): Likewise.
16659         (BUILT_IN_STPCPY): Likewise.
16660         (BUILT_IN_STRCAT): Likewise.
16661         (BUILT_IN_STRCHR): Likewise.
16662         (BUILT_IN_STRCPY): Likewise.
16663         (BUILT_IN_STRLEN): Likewise.
16664         (BUILT_IN_MEMCPY_CHK): Likewise.
16665         (BUILT_IN_MEMMOVE_CHK): Likewise.
16666         (BUILT_IN_MEMPCPY_CHK): Likewise.
16667         (BUILT_IN_MEMSET_CHK): Likewise.
16668         (BUILT_IN_STPCPY_CHK): Likewise.
16669         (BUILT_IN_STRCAT_CHK): Likewise.
16670         (BUILT_IN_STRCPY_CHK): Likewise.
16671         * calls.c (store_bounds): Likewise.
16672         (emit_call_1): Likewise.
16673         (special_function_p): Likewise.
16674         (maybe_warn_nonstring_arg): Likewise.
16675         (initialize_argument_information): Likewise.
16676         (finalize_must_preallocate): Likewise.
16677         (compute_argument_addresses): Likewise.
16678         (expand_call): Likewise.
16679         * cfgexpand.c (expand_call_stmt): Likewise.
16680         (expand_return): Likewise.
16681         (expand_gimple_stmt_1): Likewise.
16682         (pass_expand::execute): Likewise.
16683         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16684         (cgraph_node::remove): Likewise.
16685         (cgraph_node::dump): Likewise.
16686         (cgraph_node::verify_node): Likewise.
16687         * cgraph.h (chkp_function_instrumented_p): Likewise.
16688         (symtab_node::get_alias_target): Likewise.
16689         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16690         (cgraph_local_p): Likewise.
16691         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16692         (cgraph_edge::rebuild_references): Likewise.
16693         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16694         (walk_polymorphic_call_targets): Likewise.
16695         (cgraph_node::expand_thunk): Likewise.
16696         (symbol_table::output_weakrefs): Likewise.
16697         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16698         (ix86_handle_option): Likewise.
16699         * config/i386/constraints.md: Likewise.
16700         * config/i386/i386-builtin-types.def (BND): Likewise.
16701         (VOID): Likewise.
16702         (PVOID): Likewise.
16703         (ULONG): Likewise.
16704         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16705         (BDESC_FIRST): Likewise.
16706         (BDESC): Likewise.
16707         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16708         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16709         * config/i386/i386.c (enum reg_class): Likewise.
16710         (ix86_target_string): Likewise.
16711         (ix86_option_override_internal): Likewise.
16712         (ix86_conditional_register_usage): Likewise.
16713         (ix86_valid_target_attribute_inner_p): Likewise.
16714         (ix86_set_indirect_branch_type): Likewise.
16715         (ix86_set_current_function): Likewise.
16716         (ix86_function_arg_regno_p): Likewise.
16717         (init_cumulative_args): Likewise.
16718         (ix86_function_arg_advance): Likewise.
16719         (ix86_function_arg): Likewise.
16720         (ix86_pass_by_reference): Likewise.
16721         (ix86_function_value_regno_p): Likewise.
16722         (ix86_function_value_1): Likewise.
16723         (ix86_function_value_bounds): Likewise.
16724         (ix86_return_in_memory): Likewise.
16725         (ix86_setup_incoming_vararg_bounds): Likewise.
16726         (ix86_va_start): Likewise.
16727         (indirect_thunk_need_prefix): Likewise.
16728         (print_reg): Likewise.
16729         (ix86_print_operand): Likewise.
16730         (ix86_expand_call): Likewise.
16731         (ix86_output_function_return): Likewise.
16732         (reg_encoded_number): Likewise.
16733         (BDESC_VERIFYS): Likewise.
16734         (ix86_init_mpx_builtins): Likewise.
16735         (ix86_init_builtins): Likewise.
16736         (ix86_emit_cmove): Likewise.
16737         (ix86_emit_move_max): Likewise.
16738         (ix86_expand_builtin): Likewise.
16739         (ix86_builtin_mpx_function): Likewise.
16740         (ix86_get_arg_address_for_bt): Likewise.
16741         (ix86_load_bounds): Likewise.
16742         (ix86_store_bounds): Likewise.
16743         (ix86_load_returned_bounds): Likewise.
16744         (ix86_store_returned_bounds): Likewise.
16745         (ix86_class_likely_spilled_p): Likewise.
16746         (ix86_hard_regno_mode_ok): Likewise.
16747         (x86_order_regs_for_local_alloc): Likewise.
16748         (ix86_mitigate_rop): Likewise.
16749         (ix86_bnd_prefixed_insn_p): Likewise.
16750         (ix86_mpx_bound_mode): Likewise.
16751         (ix86_make_bounds_constant): Likewise.
16752         (ix86_initialize_bounds): Likewise.
16753         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16754         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16755         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16756         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16757         (TARGET_CHKP_BOUND_MODE): Likewise.
16758         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16759         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16760         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16761         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16762         * config/i386/i386.h (TARGET_MPX): Likewise.
16763         (TARGET_MPX_P): Likewise.
16764         (VALID_BND_REG_MODE): Likewise.
16765         (FIRST_BND_REG): Likewise.
16766         (LAST_BND_REG): Likewise.
16767         (enum reg_class): Likewise.
16768         (BND_REG_P): Likewise.
16769         (BND_REGNO_P): Likewise.
16770         (BNDmode): Likewise.
16771         (ADJUST_INSN_LENGTH): Likewise.
16772         * config/i386/i386.md: Likewise.
16773         * config/i386/i386.opt: Likewise.
16774         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16775         (defined): Likewise.
16776         (LINK_MPX): Likewise.
16777         (MPX_SPEC): Likewise.
16778         (LIBMPX_SPEC): Likewise.
16779         (LIBMPXWRAPPERS_SPEC): Likewise.
16780         (CHKP_SPEC): Likewise.
16781         * config/i386/predicates.md: Likewise.
16782         * dbxout.c (dbxout_type): Likewise.
16783         * doc/extend.texi: Likewise.
16784         * doc/invoke.texi: Likewise.
16785         * doc/md.texi: Likewise.
16786         * doc/tm.texi: Likewise.
16787         * doc/tm.texi.in: Likewise.
16788         * dwarf2out.c (is_base_type): Likewise.
16789         (gen_formal_types_die): Likewise.
16790         (gen_subprogram_die): Likewise.
16791         (gen_type_die_with_usage): Likewise.
16792         (gen_decl_die): Likewise.
16793         (dwarf2out_late_global_decl): Likewise.
16794         * expr.c (expand_assignment): Likewise.
16795         (emit_storent_insn): Likewise.
16796         (store_expr_with_bounds): Likewise.
16797         (store_expr): Likewise.
16798         (expand_expr_real_1): Likewise.
16799         * expr.h (store_expr_with_bounds): Likewise.
16800         * function.c (use_register_for_decl): Likewise.
16801         (struct bounds_parm_data): Likewise.
16802         (assign_parms_augmented_arg_list): Likewise.
16803         (assign_parm_find_entry_rtl): Likewise.
16804         (assign_parm_is_stack_parm): Likewise.
16805         (assign_parm_load_bounds): Likewise.
16806         (assign_bounds): Likewise.
16807         (assign_parms): Likewise.
16808         (expand_function_start): Likewise.
16809         * gcc.c (CHKP_SPEC): Likewise.
16810         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16811         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16812         (wrestrict_dom_walker::check_call): Likewise.
16813         * gimple.c (gimple_build_call_from_tree): Likewise.
16814         * gimple.h (enum gf_mask): Likewise.
16815         (gimple_call_with_bounds_p): Likewise.
16816         (gimple_call_set_with_bounds): Likewise.
16817         * gimplify.c (gimplify_init_constructor): Likewise.
16818         * ipa-cp.c (initialize_node_lattices): Likewise.
16819         (propagate_constants_across_call): Likewise.
16820         (find_more_scalar_values_for_callers_subset): Likewise.
16821         * ipa-hsa.c (process_hsa_functions): Likewise.
16822         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16823         * ipa-icf.c (sem_function::merge): Likewise.
16824         * ipa-inline.c (early_inliner): Likewise.
16825         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16826         (warn_function_cold): Likewise.
16827         (propagate_pure_const): Likewise.
16828         * ipa-ref.h (enum GTY): Likewise.
16829         * ipa-split.c (find_retbnd): Likewise.
16830         (consider_split): Likewise.
16831         (split_function): Likewise.
16832         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16833         * ipa.c (walk_polymorphic_call_targets): Likewise.
16834         (symbol_table::remove_unreachable_nodes): Likewise.
16835         (process_references): Likewise.
16836         (cgraph_build_static_cdtor_1): Likewise.
16837         * lto-cgraph.c (lto_output_node): Likewise.
16838         (output_refs): Likewise.
16839         (compute_ltrans_boundary): Likewise.
16840         (input_overwrite_node): Likewise.
16841         (input_node): Likewise.
16842         (input_cgraph_1): Likewise.
16843         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16844         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16845         (class pass_chkp_instrumentation_passes): Likewise.
16846         (make_pass_chkp_instrumentation_passes): Likewise.
16847         * passes.def: Likewise.
16848         * rtl.h (struct GTY): Likewise.
16849         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16850         * stor-layout.c (layout_type): Likewise.
16851         * symtab.c: Likewise.
16852         * target.def: Likewise.
16853         * targhooks.c (default_chkp_bound_type): Likewise.
16854         (default_chkp_bound_mode): Likewise.
16855         (default_builtin_chkp_function): Likewise.
16856         (default_chkp_function_value_bounds): Likewise.
16857         (default_chkp_make_bounds_constant): Likewise.
16858         (default_chkp_initialize_bounds): Likewise.
16859         * targhooks.h (default_chkp_bound_type): Likewise.
16860         (default_chkp_bound_mode): Likewise.
16861         (default_builtin_chkp_function): Likewise.
16862         (default_chkp_function_value_bounds): Likewise.
16863         (default_chkp_make_bounds_constant): Likewise.
16864         (default_chkp_initialize_bounds): Likewise.
16865         * toplev.c (compile_file): Likewise.
16866         (process_options): Likewise.
16867         * tree-core.h (DEF_BUILTIN): Likewise.
16868         (DEF_BUILTIN_CHKP): Likewise.
16869         * tree-inline.c (declare_return_variable): Likewise.
16870         (remap_gimple_stmt): Likewise.
16871         (copy_bb): Likewise.
16872         (initialize_inlined_parameters): Likewise.
16873         (expand_call_inline): Likewise.
16874         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16875         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16876         (make_pass_ipa_chkp_produce_thunks): Likewise.
16877         (make_pass_chkp): Likewise.
16878         (make_pass_chkp_opt): Likewise.
16879         (make_pass_chkp_instrumentation_passes): Likewise.
16880         * tree-pretty-print.c (dump_generic_node): Likewise.
16881         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16882         * tree-ssa-dce.c (propagate_necessity): Likewise.
16883         (eliminate_unnecessary_stmts): Likewise.
16884         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16885         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16886         * tree-ssa-sccvn.h: Likewise.
16887         * tree-ssa-strlen.c (get_string_length): Likewise.
16888         (valid_builtin_call): Likewise.
16889         (adjust_last_stmt): Likewise.
16890         (handle_builtin_strchr): Likewise.
16891         (handle_builtin_strcpy): Likewise.
16892         (handle_builtin_stxncpy): Likewise.
16893         (handle_builtin_memcpy): Likewise.
16894         (handle_builtin_strcat): Likewise.
16895         (strlen_check_and_optimize_stmt): Likewise.
16896         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16897         * tree-streamer-in.c: Likewise.
16898         * tree-streamer.c (record_common_node): Likewise.
16899         * tree.c (tree_code_size): Likewise.
16900         (wide_int_to_tree_1): Likewise.
16901         (type_contains_placeholder_1): Likewise.
16902         (build_common_tree_nodes): Likewise.
16903         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16904         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16905         (POINTER_BOUNDS_P): Likewise.
16906         (BOUNDED_TYPE_P): Likewise.
16907         (BOUNDED_P): Likewise.
16908         (CALL_WITH_BOUNDS_P): Likewise.
16909         (pointer_bounds_type_node): Likewise.
16910         * value-prof.c (gimple_ic): Likewise.
16911         * var-tracking.c (vt_add_function_parameters): Likewise.
16912         * varasm.c (make_decl_rtl): Likewise.
16913         (assemble_start_function): Likewise.
16914         (output_constant): Likewise.
16915         (maybe_assemble_visibility): Likewise.
16916         * varpool.c (ctor_for_folding): Likewise.
16917         * chkp-builtins.def: Remove.
16918         * ipa-chkp.c: Remove.
16919         * ipa-chkp.h: Remove.
16920         * rtl-chkp.c: Remove.
16921         * rtl-chkp.h: Remove.
16922         * tree-chkp-opt.c: Remove.
16923         * tree-chkp.c: Remove.
16924         * tree-chkp.h: Remove.
16926 2018-06-07  Carl Love  <cel@us.ibm.com>
16928         * config/rs6000/vsx.md (vextract_fp_from_shorth,
16929         vextract_fp_from_shortl): Add BE support.
16931 2018-06-07  Paul Koning  <ni1d@arrl.net>
16933         * compare-elim.c (try_merge_compare): Don't merge compare if
16934         address contains a side effect.
16935         (try_eliminate_compare): Likewise.
16937 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
16939         * config.gcc: Support "tremont".
16940         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16941         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16942         PROCESSOR_TREMONT.
16943         * config/i386/i386.c (m_TREMONT): Define.
16944         (processor_target_table): Add "tremont".
16945         (PTA_TREMONT): Define.
16946         (ix86_lea_outperforms): Add TARGET_TREMONT.
16947         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16948         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16949         and M_INTEL_GOLDMONT_PLUS.
16950         (fold_builtin_cpu): Add "tremont".
16951         (ix86_add_stmt_cost): Add TARGET_TREMONT.
16952         (ix86_option_override_internal): Add "tremont".
16953         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16954         (processor_type): Add PROCESSOR_TREMONT.
16955         * config/i386/x86-tune.def: Add m_TREMONT.
16956         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16958 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16960         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16961         symbol defined for msp430i* devices to be lower case.
16963 2018-06-07  Richard Biener  <rguenther@suse.de>
16965         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16966         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
16967         Properly wrap signed arithmetic if overflow wraps.
16969 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
16971         PR tree-optimization/69615
16972         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16973         of a cast from a same precision integral SSA_NAME in a bb dominated
16974         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
16975         cast to utype if rhs2 has already a compatible type.
16977 2018-06-07  Richard Biener  <rguenther@suse.de>
16979         PR tree-optimization/85935
16980         * graphite-scop-detection.c (find_params_in_bb): Analyze
16981         condition operands with respect to the correct loop.  Assert
16982         the analysis doesn't fail.
16984 2018-06-04  Carl Love  <cel@us.ibm.com>
16986         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16987         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
16988         as it is slightly cheaper.
16989         (first_match_or_eos_index_<mode>):
16990         Calculate index using natural element order.
16991         (first_match_index_<mode>):
16992         Calculate index using natural element order.
16993         (first_match_or_eos_index_<mode>):
16994         Calculate index using natural order.
16995         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16996         for BE and LE modes.
16997         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16998         P9V_BUILTIN_VCLZLSBB_V16QI.
16999         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17000         specific.
17002 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17004         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17005         indentation and line wrap for many prototypes.  Add missing
17006         @smallexample directives around block of prototypes for vec_xl and
17007         vec_xst.
17009 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17011         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17012         track if we pass or return IEEE 128-bit floating point.
17013         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17014         C++ mangling that is compatible with GCC 8.1.
17015         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17016         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17017         floating point types.
17018         (rs6000_function_arg_advance_1): Likewise.
17019         (rs6000_mangle_type): Optionally generate mangled names that match
17020         what GCC 8.1 generated for IEEE 128-bit floating point types.
17021         (rs6000_globalize_decl_name): If we have an external function that
17022         passes or returns IEEE 128-bit types, generate a weak reference
17023         from the mangled name used in GCC 8.1 to the current mangled
17024         name.
17025         (rs6000_init_builtins): Make __ibm128 use the long double type if
17026         long double is IBM extended double.  Make __float128 use the long
17027         double type if long double is IEEE 128-bit.
17029         PR target/85657
17030         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17031         macro for __ibm128 built-in functions.
17032         (PACK_IF): Add __ibm128 pack/unpack functions.
17033         (UNPACK_IF): Likewise.
17034         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17035         enable long double built-in functions if long double is IEEE
17036         128-bit floating point.
17037         (rs6000_invalid_builtin): Update long double built-in function
17038         error message.
17039         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17040         functions, adjust the built-in function to use the long double
17041         built-in function if __ibm128 and long double are the same type.
17042         * doc/extend.texi (PowerPC builtins): Update documention for
17043         __builtin_{,un}pack_longdouble.  Add documentation for
17044         __builtin_{,un}pack_ibm128.
17046 2018-06-06  Jim Wilson  <jimw@sifive.com>
17048         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17049         (struct machine_function): New field interrupt_mode.
17050         (riscv_handle_type_attribute): New function.  Add forward declaration.
17051         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17052         (riscv_expand_epilogue): Check interrupt_mode field.
17053         (riscv_set_current_function): Check interrupt attribute args and
17054         set interrupt_mode field.
17055         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17056         (riscv_sret, riscv_uret): New.
17057         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17058         new arguments to interrupt attribute.
17060 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17062         PR target/63177
17063         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17064         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17066 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17068         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17069         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17070         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17071         several redundant entries.
17073 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17075         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17076         type from "rtx" to "rtx_insn *".
17077         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17078         for local "call_insn", removing cast.
17079         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17081 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17083         PR tree-optimization/86066
17084         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17085         for BIT_INSERT_EXPR stores.
17087 2018-06-06  Richard Biener  <rguenther@suse.de>
17089         PR tree-optimization/86062
17090         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17091         component refs ontop
17092         of to be offsetted base.
17094 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17096         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17097         to be static and remove check on interrupt attribute name.
17099 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17101         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17102         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17104 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17106         PR target/79924
17107         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17108         second argument.
17109         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17110         Remove second argument, change how error is called.
17111         (aarch64_layout_arg): Remove second argument from
17112         aarch64_err_no_fpadvsimd call.
17113         (aarch64_init_cumulative_args): Ditto.
17114         (aarch64_gimplify_va_arg_expr): Ditto.
17115         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17117 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17119         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17120         (*simple_return_indirect_internal<mode>): Rename from
17121         simple_return_indirect_internal.  Use W mode iterator.
17122         (rstorssp): New expander.
17123         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17124         (clrssbsy): New expander.
17125         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17127 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17129         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17130         __typeof__.
17131         (cmse_check_pointed_object): Likewise.
17133 2018-06-05  Martin Liska  <mliska@suse.cz>
17135         PR gcov-profile/47618
17136         * doc/invoke.texi: Document how -fprofile-dir format
17137         is extended.
17139 2018-06-05  Richard Biener  <rguenther@suse.de>
17141         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17142         removal pretend DOM info isn't available so we do not update
17143         it and only remove edges, not dominated blocks.  Actually free
17144         DOM info in case we removed something.  Remove unreachable blocks.
17145         (mfb_keep_latches): Work with either DOM info or marked backedges.
17146         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17147         first.  Mark backedges if DOM info isn't available.
17148         (Re-)compute DOM info after cleanup_control_flow_pre.
17150 2018-06-05  Richard Biener  <rguenther@suse.de>
17152         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17153         (locus_discrim_hasher::hash): Adjust.
17154         (locus_discrim_hasher::equal): Likewise.
17155         (next_discriminator_for_locus): Work on line directly.
17156         (same_line_p): Pass in expanded locus1 as well.
17157         (assign_discriminators): Avoid redundant location expansions.
17159 2018-06-05  Richard Biener  <rguenther@suse.de>
17161         PR tree-optimization/86046
17162         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17163         if required after clearing TREE_ADDRESSABLE.
17165 2018-06-05  Richard Biener  <rguenther@suse.de>
17167         PR tree-optimization/86047
17168         * tree-ssa-loop.c (for_each_index): Glob handling of all
17169         decls and constants and really handle all of them.
17171 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17173         PR target/81497
17174         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17175         qualifier_void_pointer and qualifier_const_void_pointer.
17176         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17177         (arm_init_builtins): Handle the above.
17178         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17179         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17180         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17181         void intrinsics.
17183 2018-06-05  Martin Liska  <mliska@suse.cz>
17185         * auto-profile.c (read_autofdo_file): Do not use
17186         gcov_ctr_summary struct.
17187         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17188         * coverage.c (struct counts_entry): Likewise.
17189         (read_counts_file): Read just single summary entry.
17190         (get_coverage_counts): Use gcov_summary struct.
17191         * coverage.h (get_coverage_counts): Likewise.
17192         * gcov-dump.c (dump_working_sets): Likewise.
17193         (tag_summary): Dump just single summary.
17194         * gcov-io.c (gcov_write_summary): Write just histogram
17195         summary.
17196         (gcov_read_summary): Read just single summary.
17197         (compute_working_sets): Use gcov_summary struct.
17198         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17199         of GCOV_COUNTERS_SUMMABLE.
17200         (GCOV_COUNTERS_SUMMABLE): Remove.
17201         (GCOV_FIRST_VALUE_COUNTER): Replace with
17202         GCOV_COUNTER_V_INTERVAL.
17203         (struct gcov_ctr_summary): Remove.
17204         (struct gcov_summary): Directly use fields of former
17205         gcov_ctr_summary.
17206         (compute_working_sets): Use gcov_summary struct.
17207         * gcov.c (read_count_file): Do not use ctrs fields.
17208         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17209         struct.
17210         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17211         struct.
17212         * profile.c: Likewise.
17213         * profile.h: Likewise.
17215 2018-06-05  Martin Liska  <mliska@suse.cz>
17217         PR gcov-profile/84846
17218         * gcov.c (output_lines): Print working directory only
17219         in intermediate format.
17221 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17223         * config/s390/s390-builtin-types.def: Add void function type.
17224         * config/s390/s390-builtins.def: Use the function type for the
17225         tbeginc builtin.
17227 2018-06-04  Jim Wilson  <jimw@sifive.com>
17229         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17230         to int.
17231         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17232         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17233         handle EH_RETURN_DATA_REGNO registers properly.
17234         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17235         (riscv_expand_epilogue): Update comment.  Change argument name and
17236         type.  Update code to use new name and type.  Pass new args to
17237         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17238         EXCEPTION_RETURN.
17239         * config/riscv/riscv.md (NORMAL_RETURN): New.
17240         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17241         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17242         (eh_return): Call gen_eh_return_internal and emit barrier.
17243         (eh_return_internal): Call riscv_expand_epilogue.
17245 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17247         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17248         bit_insertion field and declare can_be_merged_into method.
17249         (merged_store_group::can_be_merged_into): New method.
17250         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17251         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17252         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17254 2018-06-04  Richard Biener  <rguenther@suse.de>
17256         PR tree-optimization/85955
17257         * builtins.c (fold_builtin_sincos): Convert pointers to
17258         destination to appropriate type before dereferencing.
17260 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17262         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17264 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17266         * expr.c (expand_expr_real_1): Force the operand into memory if
17267         its TYPE_MODE is BLKmode and if there is no integer mode for
17268         the number of bits being extracted.
17270 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17272         PR target/85832
17273         PR target/86036
17274         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17275         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17277 2018-06-04  Richard Biener  <rguenther@suse.de>
17279         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17280         (cleanup_tree_cfg_noloop): ... single caller.  Do
17281         start_recording_case_labels later.
17283 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17285         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17286         to _IMMINTRIN_H_INCLUDED.
17287         * config/i386/pconfigintrin.h: Ditto.
17288         * config/i386/waitpkgintrin.h: Ditto.
17289         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17290         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17291         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17292         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17293         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17294         waitpkgintrin.h and cldemoteintrin.h.
17296 2018-06-04  Richard Biener  <rguenther@suse.de>
17298         PR tree-optimization/86038
17299         * tracer.c (find_best_successor): Check probability for
17300         being initialized, bail out if not.
17302 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17304         PR target/86003
17305         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17306         of bits to ignore when comparing architectures.
17308 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17310         PR tree-optimization/69615
17311         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17312         maximum or minimum of the type, try to merge it also as if
17313         range1 is + [-, x - 1] or + [x + 1, -].
17315         PR c++/86025
17316         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17318 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17320         PR tree-optimization/86034
17321         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17322         the unsigned bitfield type in a bit insertion sequence if it does not
17323         have a larger precision than the bitfield size.
17324         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17326 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17328         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17330 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17332         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17333         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17334         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17335         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17337 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17339         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17340         Disable -fdelete-null-pointer-checks for ELF toolchain.
17342 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17343             Kito Cheng  <kito.cheng@gmail.com>
17345         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17346         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17347         (nds32*-*-*): Add glibc and uclibc conditions.
17348         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17349         (TARGET_EXCEPT_UNWIND_INFO): Define.
17350         * config/nds32/elf.h: New file.
17351         * config/nds32/linux.h: New file.
17352         * config/nds32/nds32-elf.opt: New file.
17353         * config/nds32/nds32-linux.opt: New file.
17354         * config/nds32/nds32-fp-as-gp.c
17355         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17356         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17357         TARGET_LINUX_ABI.
17358         (nds32_asm_file_end): Ditto.
17359         (nds32_print_operand): Ditto.
17360         (nds32_insert_attributes): Ditto.
17361         (nds32_init_libfuncs): New function.
17362         (TARGET_HAVE_TLS): Define.
17363         (TARGET_INIT_LIBFUNCS): Define.
17364         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17365         spec content.
17366         (TARGET_ELF): Apply different mcmodel setting.
17367         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17368         been migrated into elf.h and linux.h files.
17369         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17370         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17371         (mcmodel): The content has been migrated into nds32-elf.opt and
17372         nds32-linux.opt files.
17373         * config/nds32/t-elf: New file.
17374         * config/nds32/t-linux: New file.
17376 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17377             Shiva Chen  <shiva0217@gmail.com>
17379         * config/nds32/constants.md (unspec_volatile_element): Add
17380         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17381         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17382         optimization.
17383         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17384         (make_pass_nds32_fp_as_gp): Declare.
17385         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17386         optmization pass.
17387         (nds32_asm_function_end_prologue): Remove unused asm output.
17388         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17389         (nds32_asm_file_start): Output necessary fp_as_gp information.
17390         (nds32_option_override): Adjust register usage.
17391         (nds32_expand_prologue): Consider fp_as_gp situation.
17392         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17393         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17394         (epilogue): Ditto.
17395         (return): Ditto.
17396         (simple_return): Ditto.
17397         (omit_fp_begin): Output special directive for fp_as_gp.
17398         (omit_fp_end): Output special directive for fp_as_gp.
17399         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17400         mforbid-fp-as-gp): New options.
17402 2018-06-01  Mark Wielaard  <mark@klomp.org>
17404         * dwarf2out.c (dwarf2out_finish): Remove generation of
17405         DW_AT_loclists_base.
17407 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17409         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17410         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17411         (struct merged_store_group): Add bit_insertion field.
17412         (dump_char_array): Use standard hexadecimal format.
17413         (merged_store_group::merged_store_group): Set bit_insertion to false.
17414         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17415         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17416         also print the mask in the dump file.
17417         (pass_store_merging::gate): Minor tweak.
17418         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17419         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17420         stores with INTEGER_CST stores.
17421         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17422         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17423         and use it throughout.  Generate bit insertion sequences if need be.
17424         (pass_store_merging::process_store): Remove redundant condition.
17425         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17427 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17429         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17430         the 128-bit floating point types.  Fix function comment.
17432 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17434         * config/aarch64/aarch64-simd.md
17435         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17436         mnemonics.
17437         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17438         mnemonics.
17440 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17442         PR tree-optimization/85989
17443         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17444         variable.
17445         (backprop::intersect_uses): Check it when deciding whether this
17446         is a backedge reference.
17447         (backprop::process_block): Add each phi to m_visited_phis
17448         after visiting it, then clear it at the end.
17450 2018-06-01  Richard Biener  <rguenther@suse.de>
17452         * tree-vectorizer.h (vect_dr_stmt): New function.
17453         (vect_get_load_cost): Adjust.
17454         (vect_get_store_cost): Likewise.
17455         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17456         Use vect_dr_stmt instead of DR_SMTT.
17457         (vect_record_base_alignments): Likewise.
17458         (vect_calculate_target_alignment): Likewise.
17459         (vect_compute_data_ref_alignment): Likewise and make static.
17460         (vect_update_misalignment_for_peel): Likewise.
17461         (vect_verify_datarefs_alignment): Likewise.
17462         (vector_alignment_reachable_p): Likewise.
17463         (vect_get_data_access_cost): Likewise.  Pass down
17464         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17465         (vect_get_peeling_costs_all_drs): Likewise.
17466         (vect_peeling_hash_get_lowest_cost): Likewise.
17467         (vect_enhance_data_refs_alignment): Likewise.
17468         (vect_find_same_alignment_drs): Likewise.
17469         (vect_analyze_data_refs_alignment): Likewise.
17470         (vect_analyze_group_access_1): Likewise.
17471         (vect_analyze_group_access): Likewise.
17472         (vect_analyze_data_ref_access): Likewise.
17473         (vect_analyze_data_ref_accesses): Likewise.
17474         (vect_vfa_segment_size): Likewise.
17475         (vect_small_gap_p): Likewise.
17476         (vectorizable_with_step_bound_p): Likewise.
17477         (vect_prune_runtime_alias_test_list): Likewise.
17478         (vect_analyze_data_refs): Likewise.
17479         (vect_supportable_dr_alignment): Likewise.
17480         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17481         (vect_gen_prolog_loop_niters): Likewise.
17482         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17483         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17484         modify DR_STMT.
17485         (vect_recog_mask_conversion_pattern): Likewise.
17486         (vect_try_gather_scatter_pattern): Likewise.
17487         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17488         to vect_get_store_cost.
17489         (vect_get_store_cost): Get stmt_info instead of DR.
17490         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17491         (vect_get_load_cost): Get stmt_info instead of DR.
17493 2018-06-01  Richard Biener  <rguenther@suse.de>
17495         PR middle-end/86017
17496         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17497         vars wrapped in MEM_REFs.
17499 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17501         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17502         Fix subreg tests so that we only return a choice between
17503         GENERAL_REGS and FP_REGS if the original classes included both.
17505 2018-06-01  Richard Biener  <rguenther@suse.de>
17507         PR ipa/85960
17508         * tree-ssa-structalias.c (get_function_part_constraint):
17509         Handle NULL fi->decl.
17510         (find_func_aliases_for_call): Properly handle indirect
17511         fi from direct call.
17512         (find_func_clobbers): Likewise.
17513         (ipa_pta_execute): Likewise.
17514         (create_variable_info_for): For functions that are ifunc_resolver
17515         resolve to a varinfo that contains the result of the resolver call.
17516         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17517         aliases.
17519 2018-05-31  Michael Collison  <michael.collison@arm.com>
17521         * config/aarch64/aarch64.md:
17522         (*fix_to_zero_extenddfdi2): New pattern.
17523         * gcc.target/aarch64/fix_extend1.c: New testcase.
17525 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17527         PR middle-end/78809
17528         PR middle-end/83026
17529         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17530         and BUILT_IN_STRNCMP_EQ.
17531         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17532         BUILT_IN_STRNCMP_EQ.
17533         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17534         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17535         (gimple_fold_builtin): Likewise.
17536         * tree-ssa-strlen.c (compute_string_length): New function.
17537         (determine_min_obsize): New function.
17538         (handle_builtin_string_cmp): New function to handle calls to
17539         string compare functions.
17540         (strlen_optimize_stmt): Add handling to builtin string compare
17541         calls.
17542         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17543         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17544         * tree.c (build_common_builtin_nodes): Add new defines of
17545         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17547 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17549         PR target/85984
17550         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17551         functions with naked attribute.
17553 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17555         * config/i386/sse.md (avx_vec_concat<mode>):
17556         Substitute concat_tg_mode mode attribute with xtg_mode.
17557         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17558         (concat_tg_mode): Remove mode attribute.
17560 2018-05-31  Martin Sebor  <msebor@redhat.com>
17562         PR c/82063
17563         * calls.c (alloc_max_size): Correct a logic error/typo.
17564         Treat excessive arguments as infinite.  Warn for invalid arguments.
17565         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17567 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17569         PR target/85829
17570         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17571         and movx for Haswell.
17573 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17574             Cesar Philippidis  <cesar@codesourcery.com>
17576         PR middle-end/85879
17577         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17578         when emitting error on private/firstprivate reductions.
17579         * omp-low.c (lower_omp_target): Avoid reference-type processing
17580         on pointers for firstprivate clause.
17582 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17584         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17585         (st1x2): Likewise.
17586         (st1x3): Likewise.
17587         * config/aarch64/aarch64-simd.md
17588         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17589         (aarch64_ld1_x3_<mode>): Likewise
17590         (aarch64_st1x2<VALLDIF:mode>): Likewise
17591         (aarch64_st1_x2_<mode>): Likewise
17592         (aarch64_st1x3<VALLDIF:mode>): Likewise
17593         (aarch64_st1_x3_<mode>): Likewise
17594         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17595         (vld1_s8_x3): Likewise.
17596         (vld1_u16_x3): Likewise.
17597         (vld1_s16_x3): Likewise.
17598         (vld1_u32_x3): Likewise.
17599         (vld1_s32_x3): Likewise.
17600         (vld1_u64_x3): Likewise.
17601         (vld1_s64_x3): Likewise.
17602         (vld1_f16_x3): Likewise.
17603         (vld1_f32_x3): Likewise.
17604         (vld1_f64_x3): Likewise.
17605         (vld1_p8_x3): Likewise.
17606         (vld1_p16_x3): Likewise.
17607         (vld1_p64_x3): Likewise.
17608         (vld1q_u8_x3): Likewise.
17609         (vld1q_s8_x3): Likewise.
17610         (vld1q_u16_x3): Likewise.
17611         (vld1q_s16_x3): Likewise.
17612         (vld1q_u32_x3): Likewise.
17613         (vld1q_s32_x3): Likewise.
17614         (vld1q_u64_x3): Likewise.
17615         (vld1q_s64_x3): Likewise.
17616         (vld1q_f16_x3): Likewise.
17617         (vld1q_f32_x3): Likewise.
17618         (vld1q_f64_x3): Likewise.
17619         (vld1q_p8_x3): Likewise.
17620         (vld1q_p16_x3): Likewise.
17621         (vld1q_p64_x3): Likewise.
17622         (vst1_s64_x2): Likewise.
17623         (vst1_u64_x2): Likewise.
17624         (vst1_f64_x2): Likewise.
17625         (vst1_s8_x2): Likewise.
17626         (vst1_p8_x2): Likewise.
17627         (vst1_s16_x2): Likewise.
17628         (vst1_p16_x2): Likewise.
17629         (vst1_s32_x2): Likewise.
17630         (vst1_u8_x2): Likewise.
17631         (vst1_u16_x2): Likewise.
17632         (vst1_u32_x2): Likewise.
17633         (vst1_f16_x2): Likewise.
17634         (vst1_f32_x2): Likewise.
17635         (vst1_p64_x2): Likewise.
17636         (vst1q_s8_x2): Likewise.
17637         (vst1q_p8_x2): Likewise.
17638         (vst1q_s16_x2): Likewise.
17639         (vst1q_p16_x2): Likewise.
17640         (vst1q_s32_x2): Likewise.
17641         (vst1q_s64_x2): Likewise.
17642         (vst1q_u8_x2): Likewise.
17643         (vst1q_u16_x2): Likewise.
17644         (vst1q_u32_x2): Likewise.
17645         (vst1q_u64_x2): Likewise.
17646         (vst1q_f16_x2): Likewise.
17647         (vst1q_f32_x2): Likewise.
17648         (vst1q_f64_x2): Likewise.
17649         (vst1q_p64_x2): Likewise.
17650         (vst1_s64_x3): Likewise.
17651         (vst1_u64_x3): Likewise.
17652         (vst1_f64_x3): Likewise.
17653         (vst1_s8_x3): Likewise.
17654         (vst1_p8_x3): Likewise.
17655         (vst1_s16_x3): Likewise.
17656         (vst1_p16_x3): Likewise.
17657         (vst1_s32_x3): Likewise.
17658         (vst1_u8_x3): Likewise.
17659         (vst1_u16_x3): Likewise.
17660         (vst1_u32_x3): Likewise.
17661         (vst1_f16_x3): Likewise.
17662         (vst1_f32_x3): Likewise.
17663         (vst1_p64_x3): Likewise.
17664         (vst1q_s8_x3): Likewise.
17665         (vst1q_p8_x3): Likewise.
17666         (vst1q_s16_x3): Likewise.
17667         (vst1q_p16_x3): Likewise.
17668         (vst1q_s32_x3): Likewise.
17669         (vst1q_s64_x3): Likewise.
17670         (vst1q_u8_x3): Likewise.
17671         (vst1q_u16_x3): Likewise.
17672         (vst1q_u32_x3): Likewise.
17673         (vst1q_u64_x3): Likewise.
17674         (vst1q_f16_x3): Likewise.
17675         (vst1q_f32_x3): Likewise.
17676         (vst1q_f64_x3): Likewise.
17677         (vst1q_p64_x3): Likewise.
17679 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17681         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17682         user_label_prefix to name.
17684         * tree-core.h: Update comment about the format of NAME string
17685         passed to handler in attribute_spec.
17687         * config/msp430/msp430.md: Remove erroneous subreg expression from
17688         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17689         zero_extend{q,h}isi2.
17691 2018-05-30  Borislav Petkov  <bp@suse.de>
17693         * doc/extend.texi: Document some architecture specific
17694         constraints and sort entries.
17696 2018-05-30  Martin Sebor  <msebor@redhat.com>
17698         PR middle-end/85369
17699         * builtins.c (expand_builtin_stpcpy_1): New function.
17700         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17701         only if the former succeeds.
17703 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17705         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17706         in saphira.
17708 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17710         * doc/invoke.texi (-flinker-output): Document
17712 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17714         * passes.c (ipa_write_summaries): Only modify statements if body
17715         is in memory.
17716         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17717         incrementally linking.
17718         (ipa_passes): Likewise.
17719         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17720         pass down resolution info.
17721         * common.opt (flag_incremental_link): Update info.
17722         * gcc.c (plugin specs): Turn flinker-output=* to
17723         -plugin-opt=-linker-output-known
17724         * toplev.c (compile_file): Also cut compilation when doing incremental
17725         link.
17726         * flag-types. (enum lto_partition_model): Add
17727         LTO_LINKER_OUTPUT_NOLTOREL.
17728         (invoke.texi): Add -flinker-output docs.
17729         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17730         link same way as WPA; do not stream in dead initializers.
17732         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17733         darf2out_register_external_decl): Support incremental link.
17735 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17737         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17739 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17741         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17742         it down to simple_object_copy_lto_debug_sections.
17743         (run_gcc): Determine incremental LTO link time and configure
17744         lto1 into non-wpa mode, disable renaming of debug sections.
17746 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17748         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17749         descriptions of various incorrectly documented functions.
17751 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17753         Revert:
17754         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17755         address check not strict.
17757 2018-05-30  Richard Biener  <rguenther@suse.de>
17759         PR tree-optimization/85964
17760         * tracer.c (better_p): Drop initialized count check, we only
17761         call the function with initialized counts now.
17762         (find_best_successor): Do find a best edge if one
17763         has uninitialized count.
17764         (find_best_predecessor): Likewise.  Do BB frequency check only
17765         if count is initialized.
17767 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17769         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17770         (aarch64_ldrstr_offset_compare): New.
17771         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17772         load/store orderings.
17773         (aarch64_gen_adjusted_ldpstp): Likewise.
17775 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17777         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17778         Check for subset of GENERAL_REGS and FP_REGS.
17779         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17780         r=w alternative.
17782 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17784         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17785         and wi::to_poly_offset.  Add the current offset and then check
17786         whether the sum fits, rather than using an unchecked addition of
17787         a checked term.  Check for a shwi rather than a uhwi.
17788         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17789         (store_constructor): Use poly_int_tree_p.
17790         (expand_expr_real_1): Likewise.
17791         * function.c (assign_temp): Likewise.
17792         * fold-const.c (const_binop): Use poly_int_tree_p and
17793         wi::to_poly_offset.
17794         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17795         division.
17796         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17797         to_poly_offset for MEM offsets.
17798         * ipa-icf.c (sem_variable::equals): Likewise.
17799         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17800         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17801         wi::to_poly_offset for BIT_FIELD_REF offsets.
17802         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17803         wi::to_poly_offset.
17804         * var-tracking.c (emit_note_insn_var_location): Use
17805         tree_to_poly_uint64.
17807 2018-05-29  Jim Wilson  <jimw@sifive.com>
17809         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17811 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17813         PR target/85950
17814         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17815         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17816         sequence.
17817         (sse4_1_round<mode>2): Use nonimmediate_operand
17818         for operand 1 predicate.
17820 2018-05-29  Martin Sebor  <msebor@redhat.com>
17821             Richard Biener  <rguenther@suse.de>
17823         PR testsuite/85888
17824         * calls.c (get_size_range): Call determine_value_range instead
17825         of get_value_range..
17826         * tree-vrp.h (determine_value_range): Declared new function.
17827         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17829 2018-05-29  Richard Biener  <rguenther@suse.de>
17831         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17832         sure to use non-pattern stmts for get_earlier_stmt arguments.
17833         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17834         called on pattern stmts.
17835         (get_later_stmt): Likewise.
17837 2018-05-29  Martin Liska  <mliska@suse.cz>
17839         PR gcov-profile/85759
17840         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17841         env variables.
17843 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17845         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17846         VEC_UNPACK_*_EXPR.
17847         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17848         VEC_PACK_*_EXPR.
17850         PR target/85918
17851         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17852         VEC_PACK_FLOAT_EXPR): New tree codes.
17853         * tree-pretty-print.c (op_code_prio): Handle
17854         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17855         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17856         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17857         * tree-inline.c (estimate_operator_cost): Likewise.
17858         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17859         * fold-const.c (const_binop): Likewise.
17860         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17861         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17862         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17863         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17864         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17865         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17866         * expr.c (expand_expr_real_2): Likewise.
17867         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17868         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17869         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17870         optabs.
17871         * optabs.c (expand_widen_pattern_expr): For
17872         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17873         sign from result type rather than operand's type.
17874         (expand_binop_directly): For vec_packu_float_optab and
17875         vec_packs_float_optab allow result type to be different from operand's
17876         type.
17877         * optabs-tree.c (optab_for_tree_code): Handle
17878         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17879         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
17880         * tree-vect-generic.c (expand_vector_operations_1):  Handle
17881         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17882         VEC_PACK_FLOAT_EXPR.
17883         * tree-vect-stmts.c (supportable_widening_operation): Handle
17884         FIX_TRUNC_EXPR.
17885         (supportable_narrowing_operation): Handle FLOAT_EXPR.
17886         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17887         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17888         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
17889         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17890         mode attributes.
17891         (vec_pack<floatprefix>_float_<mode>): New expander.
17892         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17893         attributes.
17894         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17895         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17896         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17897         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17898         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17899         Document.
17900         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17901         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17902         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17903         VEC_PACK_FLOAT_EXPR): Document.
17905 2018-05-29  Richard Biener  <rguenther@suse.de>
17907         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17908         member.
17909         (stmt_vec_info_vec): Make pointer.
17910         (init_stmt_vec_info_vec): Remove.
17911         (free_stmt_vec_info_vec): Likewise.
17912         (set_stmt_vec_info_vec): New function.
17913         (free_stmt_vec_infos): Likewise.
17914         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17915         (set_vinfo_for_stmt): Likewise.
17916         (get_earlier_stmt): Likewise.
17917         (get_later_stmt): Likewise.
17918         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17919         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17920         (vec_info::~vec_info): Free stmt_vec_infos.
17921         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17922         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17923         (pass_slp_vectorize::execute): Likewise.
17924         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17925         (free_stmt_vec_info_vec): Likewise.
17926         (set_stmt_vec_info_vec): New function.
17927         (free_stmt_vec_infos): Likewise.
17928         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17929         the global stmt_vec_info_vec.
17930         * tree-parloops.c (gather_scalar_reductions): Use
17931         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17932         vector.
17934 2018-05-29  Richard Biener  <rguenther@suse.de>
17936         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17938 2018-05-29  Martin Liska  <mliska@suse.cz>
17939             David Malcolm  <dmalcolm@redhat.com>
17941         * vec.c (test_reverse): New.
17942         (vec_c_tests): Add new test.
17943         * vec.h (vl_ptr>::reverse): New function.
17945 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
17947         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17949         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17950         and later.
17952 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17954         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17956 2018-05-28  Richard Biener  <rguenther@suse.de>
17958         PR tree-optimization/85933
17959         * tree-vect-data-refs.c (vect_record_base_alignments): Only
17960         look at stmts marked as vectorizable.
17962 2018-05-28  Richard Biener  <rguenther@suse.de>
17964         PR tree-optimization/85934
17965         * tree-vect-generic.c (expand_vector_operations_1): Hoist
17966         vector boolean check before scalar optimization.
17968 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
17970         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17971         for armv5te.
17973 2018-05-28  Mark Wielaard  <mark@klomp.org>
17975         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17976         if it is an expression containing a minus sign.
17978 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
17980         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17982 2018-05-27  Paul Koning  <ni1d@arrl.net>
17984         * config/pdp11/pdp11.md (truncsihi2): Remove.
17986 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
17987             Chung-Ju Wu  <jasonwucj@gmail.com>
17989         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17990         implementation.
17991         (unaligned_store_dw): Ditto.
17992         * config/nds32/nds32-memory-manipulation.c
17993         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17994         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17995         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17996         (emit_setmem_word_loop): Rename to ...
17997         (emit_setmem_doubleword_loop): ... this.
17998         (nds32_gen_dup_4_byte_to_word_value): New function.
17999         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18000         (nds32_expand_setmem_loop): Refine implementation.
18001         (nds32_expand_setmem_loop_v3m): Ditto.
18002         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18003         pattern.
18005 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18007         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18009 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18011         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18012         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18013         machine->attr_no_prologue_p.
18014         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18015         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18016         (nds32_expand_epilogue): Consider attr_naked_p.
18017         (nds32_expand_epilogue_v3pop): Likewise.
18018         (nds32_can_use_return_insn): Likewise.
18019         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18020         attr_no_prologue_p fields.
18021         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18023 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18025         PR target/85918
18026         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18027         attributes.
18028         * config/i386/sse.md
18029         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18030         Rename to ...
18031         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18032         ... this.
18033         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18034         Rename to ...
18035         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18036         ... this.
18037         (*<floatsuffix>floatv2div2sf2): Rename to ...
18038         (*float<floatunssuffix>v2div2sf2): ... this.
18039         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18040         (float<floatunssuffix>v2div2sf2_mask): ... this.
18041         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18042         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18043         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18044         to ...
18045         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18046         ... this.
18047         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18048         Rename to ...
18049         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18050         ... this.
18051         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18052         Rename to ...
18053         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18054         ... this.
18055         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18056         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18057         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18058         gen_ufix_truncv8dfv8si2.
18059         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18060         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18061         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18062         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18063         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18064         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18065         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18066         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18068 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18070         PR target/85900
18071         PR target/85345
18072         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18074 2018-05-25  Jim Wilson  <jimw@sifive.com>
18076         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18077         * config/riscv/riscv.c (struct machine_function): Add
18078         interrupt_handler_p and attribute_checked_p fields.
18079         (riscv_attribute_table): Add interrupt.
18080         (riscv_interrupt_type_p): New.
18081         (riscv_save_reg_p): Save extra regs for interrupt handler.
18082         (riscv_use_save_libcall): Return false  for interrupt handler.
18083         (riscv_first_stack_step): Add forward declaration.
18084         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18085         for interrupt handler with large frame.  Use it for saved reg list.
18086         (riscv_expand_prologue): Move flag_stack_usage_info support to
18087         eliminate duplication.
18088         (riscv_expand_epilogue): Generate mret for interrupt handler.
18089         (riscv_epilogue_uses): New.
18090         (riscv_can_use_return_insn): Return false for interrupt handler.
18091         (riscv_function_ok_for_sibcall): Likewise.
18092         (riscv_set_current_function): Add interrupt handler support.
18093         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18094         * config/riscv/riscv.md (UNSPECV_MRET): New.
18095         (GP_REGNUM): New.
18096         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18097         (riscv_mret): New.
18098         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18100 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18102         PR tree-optimization/85712
18103         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18104         this candidate has already been replaced in-situ by a copy.
18106 2018-05-25  Jason Merrill  <jason@redhat.com>
18108         PR c++/80485 - inline function non-zero address.
18109         * symtab.c (nonzero_address): Check DECL_COMDAT.
18111 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18113         PR target/83628
18114         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18115         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18116         extension of SImode operation.  Use const123_operand predicate.
18117         (*saddsi_1): Remove.
18118         (*saddl_se_1): Ditto.
18119         (*ssubsi_1): Ditto.
18120         (*ssubl_se_1): Ditto.
18121         * config/alpha/predicates.md (const123_operand): New predicate.
18122         * config/alpha/constraints.md (P): Use IN_RANGE.
18124 2018-05-25  Richard Biener  <rguenther@suse.de>
18126         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18127         defaulted to true.
18128         (ref_maybe_used_by_stmt_p): Likewise.
18129         (stmt_may_clobber_ref_p): Likewise.
18130         (stmt_may_clobber_ref_p_1): Likewise.
18131         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18132         and pass it along.
18133         (ref_maybe_used_by_stmt_p): Likewise.
18134         (stmt_may_clobber_ref_p): Likewise.
18135         (stmt_may_clobber_ref_p_1): Likewise.
18136         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18137         the alias oracle to disambiguate DRs with stmts DR analysis
18138         couldn't handle.
18139         (vect_analyze_data_refs): Do not give up on not analyzable
18140         DRs for BB vectorization.  Remove code truncating the dataref
18141         vector.
18143 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18145         PR target/85832
18146         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18147         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18148         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18150 2018-05-25  Richard Biener  <rguenther@suse.de>
18152         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18153         function, combining stmt data ref gathering and fatal analysis
18154         parts.
18155         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18156         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18157         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18158         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18159         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18161 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18163         PR tree-optimization/85720
18164         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18165         SCC if all partitions are builtins.
18166         (version_loop_by_alias_check): New parameter.  Generate cancelable
18167         runtime alias check if all partitions are builtins.
18168         (distribute_loop): Update call to above function.
18170 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18172         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18173         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18174         (parm_default_def_partition_arg): Ditto.
18175         (set_parm_default_def_partition): Ditto.
18176         (get_parm_default_def_partitions): Ditto and make it static.
18177         (get_undefined_value_partitions): Ditto and make it static.
18178         (remove_ssa_form): Refactor call to init_var_map here.
18179         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18180         computation for loop region.
18181         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18182         (register_default_def): Delete.
18183         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18184         (parm_default_def_partition_arg): Ditto.
18185         (set_parm_default_def_partition): Ditto.
18186         (get_parm_default_def_partitions): Ditto and make it static.
18187         (get_undefined_value_partitions): Ditto and make it static.
18188         (coalesce_with_default, coalesce_with_default): Update comment.
18189         (create_coalesce_list_for_region): New func factored out from
18190         create_outofssa_var_map.
18191         (populate_coalesce_list_for_outofssa): New func factored out from
18192         create_outofssa_var_map and coalesce_ssa_name.
18193         (create_outofssa_var_map): Delete.
18194         (coalesce_ssa_name): Refactor to support live range computation.
18195         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18196         (get_parm_default_def_partitions): Delete.
18197         (get_undefined_value_partitions): Ditto.
18198         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18199         computation for loop region.
18200         (new_tree_live_info, loe_visit_block): Ditto.
18201         (live_worklist, set_var_live_on_entry): Ditto.
18202         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18203         * tree-ssa-live.h (struct _var_map): New fields.
18204         (init_var_map): Change decl.
18205         (region_contains_p): New.
18207 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18209         * tree-ssa-live.h (live_merge_and_clear): Delete.
18211 2018-05-25  Richard Biener  <rguenther@suse.de>
18213         PR c++/85912
18214         * tree-dump.c (dequeue_and_dump): Remove access to removed
18215         operand 2 of a SWITCH_EXPR.
18217 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18219         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18220         multiplication and division.
18221         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18222         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18223         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18224         (cond_udiv_optab, cond_umod_optab): New optabs.
18225         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18226         (IFN_COND_RDIV): New internal functions.
18227         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18228         TRUNC_MOD_EXPR and RDIV_EXPR.
18229         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18230         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18231         New unspecs.
18232         (SVE_INT_BINARY): Include mult.
18233         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18234         (optab, sve_int_op): Handle mult.
18235         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18236         UNSPEC_COND_DIV.
18237         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18238         for SVE_INT_BINARY_SD.
18240 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18242         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18243         (optab, sve_int_op): Handle div and udiv.
18244         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18245         for SVE_INT_BINARY_SD.
18246         (*<optab><mode>3): New insn for the same.
18248 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18250         * tree-vect-patterns.c: Include predict.h.
18251         (vect_recog_divmod_pattern): Restrict check for division support
18252         to when optimizing for size.
18254 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18256         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18257         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18258         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18259         (gimple_match_op::set_op): Likewise.
18260         (gimple_resimplify4): Declare.
18261         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18262         (expr::gen_transform): Likewise.
18263         (decision_tree::gen): Generate a simplification routine for 4 operands.
18264         * gimple-match-head.c (gimple_simplify): Add an overload for
18265         4 operands.  In the top-level function, handle up to 4 call
18266         arguments and call gimple_resimplify4.
18267         (gimple_resimplify4): New function.
18268         (build_call_internal): Pass a fourth operand.
18269         (maybe_push_to_seq): Likewise.
18270         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18271         Fold VEC_COND_EXPRs of an operation and a default value into
18272         an IFN_COND_* function if possible.
18273         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18274         New unspecs.
18275         (SVE_COND_FP_BINARY): Include them.
18276         (optab, sve_fp_op): Handle them.
18277         (SVE_INT_BINARY_REV): New code iterator.
18278         (SVE_COND_FP_BINARY_REV): New int iterator.
18279         (commutative): New int attribute.
18280         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18281         Declare.
18282         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18283         function.
18284         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18285         (*cond_<optab><mode>): New patterns for reversed operands.
18287 2018-05-25  Richard Biener  <rguenther@suse.de>
18289         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18290         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18291         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18292         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18293         * tree-vect-data-refs.c (everywhere): Adjust users.
18294         * tree-vect-loop.c (everywhere): Likewise.
18295         * tree-vect-slp.c (everywhere): Likewise.
18296         * tree-vect-stmts.c (everywhere): Likewise.
18297         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18299 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18301         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18302         Rename to...
18303         (gcc_cv_as_section_exclude): ... this.
18304         Try Solaris as #exclude syntax.
18305         * configure: Regenerate.
18306         * config.in: Regenerate.
18307         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18308         SECTION_EXCLUDE.
18309         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18310         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18312         * varasm.c (default_elf_asm_named_section): Don't check if
18313         HAVE_GAS_SECTION_EXCLUDE is defined.
18315 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18317         * doc/md.texi: Update the documentation of the cond_* optabs
18318         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18319         Describe the scalar case too.
18320         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18321         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18322         instead of 2.
18323         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18324         (get_conditional_internal_fn): Update comment.
18325         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18326         accumulator value as a final argument to conditional functions.
18327         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18328         a define_expand and add an "else" operand.  Assert for now that
18329         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18330         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18331         (*cond_<optab><mode>): New patterns.
18332         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18333         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18334         (UNSPEC_COND_EOR): Delete.
18335         (optab): Remove associated mappings.
18336         (SVE_INT_BINARY): New code iterator.
18337         (sve_int_op): Remove int attribute and add "minus" to the code
18338         attribute.
18339         (SVE_COND_INT_OP): Delete.
18340         (SVE_COND_FP_OP): Rename to...
18341         (SVE_COND_FP_BINARY): ...this.
18343 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18345         * optabs.c (can_reuse_operands_p): New function.
18346         (maybe_legitimize_operands): Try to reuse the results for
18347         earlier operands.
18349 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18351         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18352         Add {q} suffix to insn mnemonic.
18354 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18356         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18357         (msp430_warn_func_return): New.
18359 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18361         * fold-const.c (tree_nonzero_bits): New function.
18362         * fold-const.h (tree_nonzero_bits): Likewise.
18363         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18364         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18366 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18368         PR target/85900
18369         PR target/85345
18370         * varasm.c (assemble_alias): Check ifunc_resolver only on
18371         FUNCTION_DECL.
18373 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18375         PR target/85903
18376         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18377         when memory input operand is handled.
18379 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18381         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18382         global.
18383         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18385 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18387         * match.pd: Delay FMA folds until after vectorization.
18389 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18391         PR target/83009
18392         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18393         address check not strict.
18395 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18397         * gimple-match.h (gimple_match_op): New class.
18398         (mprts_hook): Replace parameters with a gimple_match_op *.
18399         (maybe_build_generic_op): Likewise.
18400         (gimple_simplified_result_is_gimple_val): Replace parameters with
18401         a const gimple_match_op *.
18402         (gimple_simplify): Replace code_helper * and tree * parameters with
18403         a gimple_match_op * parameter.
18404         (gimple_resimplify1): Replace code_helper *, tree and tree *
18405         parameters with a gimple_match_op * parameter.
18406         (gimple_resimplify2): Likewise.
18407         (gimple_resimplify3): Likewise.
18408         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18409         parameters with a gimple_match_op * parameter.
18410         * gimple-match-head.c (gimple_simplify): Change prototypes of
18411         auto-generated functions to take a gimple_match_op * instead of
18412         separate code_helper * and tree * parameters.  Make the same
18413         change in the top-level overload and update calls to the
18414         gimple_resimplify routines.  Update calls to the auto-generated
18415         functions and to maybe_push_res_to_seq in the publicly-facing
18416         operation-specific gimple_simplify overloads.
18417         (gimple_match_op::MAX_NUM_OPS): Define.
18418         (gimple_resimplify1): Replace rcode and ops with a single res_op
18419         parameter.  Update call to gimple_simplify.
18420         (gimple_resimplify2): Likewise.
18421         (gimple_resimplify3): Likewise.
18422         (mprts_hook): Replace parameters with a gimple_match_op *.
18423         (maybe_build_generic_op): Likewise.
18424         (build_call_internal): Replace type, nargs and ops with
18425         a gimple_match_op *.
18426         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18427         with a single gimple_match_op *.  Update calls to mprts_hook,
18428         build_call_internal and gimple_simplified_result_is_gimple_val.
18429         Factor out code that is common to the tree_code and combined_fn cases.
18430         * genmatch.c (expr::gen_transform): Replace tem_code and
18431         tem_ops with a gimple_match_op called tem_op.  Update calls
18432         to the gimple_resimplify functions and maybe_push_res_to_seq.
18433         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18434         res_ops.  Update call to the gimple_resimplify functions.
18435         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18436         (decision_tree::gen): Make the functions take a gimple_match_op *
18437         called res_op instead of separate res_code and res_ops parameters.
18438         Update call accordingly.
18439         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18440         and ops with a single res_op parameter.  Update calls to
18441         maybe_build_generic_op and maybe_push_res_to_seq.
18442         (fold_stmt_1): Update calls to gimple_simplify and
18443         replace_stmt_with_simplification.
18444         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18445         and gimple_simplified_result_is_gimple_val.
18446         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18447         gimple_simplify.
18448         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18449         with a gimple_match_op *.
18450         (vn_nary_build_or_lookup): Likewise.  Update call to
18451         vn_nary_build_or_lookup_1.
18452         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18453         gimple_match_op *.  Update calls to the gimple_resimplify routines
18454         and to gimple_simplified_result_is_gimple_val.
18455         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18456         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18457         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18458         (visit_nary_op): Likewise.
18459         (visit_reference_op_load): Likewise.
18461 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18463         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18464         modifier for printing the step amount.
18466 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18468         PR target/78849
18469         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18470         types.
18472 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18474         * doc/sourcebuild.texi (Endianness): New subsubsection.
18476 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18478         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18479         <prefetch_dynamic_strides>: New const bool field.
18480         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18481         prefetch_dynamic_strides.
18482         (exynosm1_prefetch_tune): Likewise.
18483         (thunderxt88_prefetch_tune): Likewise.
18484         (thunderx_prefetch_tune): Likewise.
18485         (thunderx2t99_prefetch_tune): Likewise.
18486         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18487         false.
18488         (aarch64_override_options_internal): Update to set
18489         PARAM_PREFETCH_DYNAMIC_STRIDES.
18490         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18491         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18492         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18493         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18494         prefetch-dynamic-strides setting.
18496 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18498         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18499         <minimum_stride>: New const int field.
18500         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18501         minimum_stride field defaulting to -1.
18502         (exynosm1_prefetch_tune): Likewise.
18503         (thunderxt88_prefetch_tune): Likewise.
18504         (thunderx_prefetch_tune): Likewise.
18505         (thunderx2t99_prefetch_tune): Likewise.
18506         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18507         <default_opt_level>: Set to 3.
18508         (aarch64_override_options_internal): Update to set
18509         PARAM_PREFETCH_MINIMUM_STRIDE.
18510         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18511         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18512         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18513         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18514         stride is constant and is below the minimum stride threshold.
18516 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18518         * config/arm/arm-cpus.in (mode26): Delete.
18519         (armv4): Delete mode26 reference.
18520         * config/arm/arm.c (arm_configure_build_target): Delete use of
18521         isa_bit_mode26.
18523 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18525         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18526         New insn pattern.
18527         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18528         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18529         for non-SSE modes.
18530         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18531         (floatunsdidf2): Ditto.
18533 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18535         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18536         (fixuns_trunc<mode>si2_avx512f): Ditto.
18537         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18538         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18539         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18541 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18543         PR rtl-optimization/79985
18544         * df-scan.c (df_insn_refs_collect): Remove special case for
18545         global registers and asm statements.
18547 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18549         * extend.texi (Global Register Variables): Rewrite the bullet list.
18550         Note that the register is available for allocation. Note that access
18551         via inline asm must use constraints. Add note about async-signal
18552         handlers. Remove paragraph about automagic register selection.
18554 2018-05-23  Richard Biener  <rguenther@suse.de>
18556         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18557         of fixed offset from memset VN.
18559 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18561         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18562         first_interp field.
18563         (alloc_cand_and_find_basis): Initialize first_interp field.
18564         (slsr_process_mul): Modify first_interp field.
18565         (slsr_process_add): Likewise.
18566         (slsr_process_cast): Modify first_interp field for each new
18567         interpretation.
18568         (slsr_process_copy): Likewise.
18569         (dump_candidate): Dump first_interp field.
18570         (replace_mult_candidate): Process all interpretations, not just
18571         subsequent ones.
18572         (replace_rhs_if_not_dup): Likewise.
18573         (replace_one_candidate): Likewise.
18575 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18577         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18578         Add new boolean.
18579         (aarch64_needs_frame_chain): New function.
18580         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18582 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18584         PR target/84882
18585         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18586         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18587         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18588         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18589         as true for strict-align.
18590         (aarch64_can_inline_p): Perform checks even when callee has no
18591         attributes to check for strict alignment.
18592         * doc/extend.texi (AArch64 Function Attributes): Document
18593         no-strict-align.
18594         * doc/invoke.texi: (AArch64 Options): Likewise.
18596 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18598         PR tree-optimization/85853
18599         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18600         the handling of the root of the node to...
18601         (vect_slp_analyze_node_operations_1): ...this new function,
18602         and run the whole thing with the child nodes' def types
18603         set according to their SLP node's def type.
18605 2018-05-23  Richard Biener  <rguenther@suse.de>
18607         PR middle-end/85874
18608         * tree-data-ref.c (create_runtime_alias_checks): Defer
18609         and ignore overflow warnings.
18611 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18613         PR tree-optimization/85822
18614         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18615         constants.
18617 2018-05-23  Richard Biener  <rguenther@suse.de>
18619         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18620         memset constants via native_interpret_expr.
18622 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18624         PR target/85345
18625         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18626         attribute.
18627         (cgraph_node::create_alias): Likewise.
18628         (cgraph_node::get_availability): Check ifunc_resolver instead
18629         of looking up ifunc attribute.
18630         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18631         * varasm.c (do_assemble_alias): Likewise.
18632         (assemble_alias): Likewise.
18633         (default_binds_local_p_3): Likewise.
18634         * cgraph.h (cgraph_node): Add ifunc_resolver.
18635         (cgraph_node::only_called_directly_or_aliased_p): Return false
18636         for IFUNC resolver.
18637         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18638         attribute.
18639         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18640         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18641         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18642         instead of looking up ifunc attribute.
18644 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18646         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18648 2018-05-22  Martin Sebor  <msebor@redhat.com>
18650         PR middle-end/85359
18651         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18652         only when expasion succeeds.
18653         (expand_builtin_strcmp): Same.
18654         (expand_builtin_strncmp): Same.
18656 2018-05-22  Martin Sebor  <msebor@redhat.com>
18658         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18660 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18661             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18663         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18664         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18665         code.
18666         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18667         Add check for legitimate_address.
18668         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18669         (aarch64_swap_ldrstr_operands): New.
18670         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18671         Define prototype.
18673 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18674             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18676         * config/aarch64/aarch64.md: New patterns to generate stp
18677         and ldp.
18678         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18679         single words and double words.
18680         (load_pair_sw, load_pair_dw): Likewise.
18681         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18682         Delete.
18683         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18684         Delete.
18685         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18686         for different mode ldpstp and add peephole for merged zero stores.
18687         Likewise for loads.
18688         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18689         Add size check.
18690         (aarch64_gen_store_pair): Rename calls to match new patterns.
18691         (aarch64_gen_load_pair): Rename calls to match new patterns.
18692         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18693         (load_pair<DREG:mode><DREG2:mode>): ... This.
18694         (store_pair<mode>): Rename to...
18695         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18696         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18697         New mode iterators.
18698         (V_INT_EQUIV): Handle SImode.
18699         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18700         New predicate.
18702 2018-05-22  Martin Sebor  <msebor@redhat.com>
18704         PR c/85623
18705         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18706         or ajust the presumed bound on an operation to avoid unnecessary
18707         warnings.
18709 2018-05-22  Martin Sebor  <msebor@redhat.com>
18711         PR tree-optimization/85826
18712         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18713         assuming that a DECL necesarily has a constant size.
18715 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18717         PR middle-end/85862
18718         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18720 2018-05-22  Richard Biener  <rguenther@suse.de>
18722         PR tree-optimization/85834
18723         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18724         non-constant and non-zero memset arguments.
18726 2018-05-22  Martin Liska  <mliska@suse.cz>
18728         PR ipa/85607
18729         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18731 2018-05-22  Richard Biener  <rguenther@suse.de>
18733         PR tree-optimization/85863
18734         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18735         comparisons when vectype is specified.
18736         (vectorizable_condition): Do not specify vectype for
18737         vect_is_simple_cond when SLP vectorizing.
18739 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18741         PR target/85657
18742         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18743         define __ibm128 as long double.
18744         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18745         as a distinct type when IEEE 128-bit support is enabled.
18746         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18747         128-bit types to use the correct functions.
18748         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18749         convert between 128-bit floating point types that have different
18750         modes but the same representation, instead of using gen_lowpart to
18751         makean alias.
18752         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18753         KFmode.
18754         (IFKF_reg): New attributes to give the register constraints for
18755         IFmode and KFmode.
18756         (extend<mode>tf2_internal): New insns to mark an explicit
18757         conversion between 128-bit floating point types that have a
18758         different mode but share the same representation.
18760 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18762         PR tree-optimization/85814
18763         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18764         a null return from get_strinfo when unsharing the next
18765         strinfo in the chain.
18767 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18769         PR gcc/84923
18770         * varasm.c (weak_finish): Clean up weak_decls.
18772 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18774         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18775         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18776         UNSPEC_UADALP values.
18777         * config/aarch64/iterators.md (ABAL): New int iterator.
18778         (ABDL2): Likewise.
18779         (ADALP): Likewise.
18780         (sur): Add mappings for the above.
18781         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18782         New define_insn.
18783         (aarch64_<sur>abal<mode>_4): Likewise.
18784         (aarch64_<sur>adalp<mode>_3): Likewise.
18785         (<sur>sadv16qi): New define_expand.
18787 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18789         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18790         (*movdf_internal): Ditto.
18791         (*rcpsf2_sse): Ditto.
18792         (*rsqrtsf2_sse): Ditto.
18793         (*sqrt<mode>2_sse): Ditto.
18795 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18797         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18798         eor3q<mode>4.
18799         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18800         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18801         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18802         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18803         vbcaxq_s64): New.
18804         * config/aarch64/arm_neon.h: Likewise.
18805         * config/aarch64/iterators.md (VQ_I): New.
18807 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18809         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18810         arc*-*-linux*.
18811         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18812         MULTILIB_DIRNAMES
18814 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18816         * config/nds32/constraints.md (S): New constraint.
18817         * config/nds32/nds32.md (call_internal): Use constraint S.
18818         (call_value_internal): Likewise.
18819         (sibcall_internal): Likewise.
18820         (sibcall_value_internal): Likewise.
18822 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18823             Chung-Ju Wu  <jasonwucj@gmail.com>
18825         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18826         into consideration.
18828 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18829             Chung-Ju Wu  <jasonwucj@gmail.com>
18831         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18832         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18833         (nds32_rtx_costs_impl): Simplify.
18834         (nds32_address_cost_impl): Simplify.
18835         (nds32_init_rtx_costs): New function.
18836         (nds32_rtx_costs_speed_prefer): Likewise.
18837         (nds32_rtx_costs_size_prefer): Likewise.
18838         (nds32_address_cost_speed_prefer): Likewise.
18839         (nds32_address_cost_speed_fwprop): Likewise.
18840         (nds32_address_cost_size_prefer): Likewise.
18841         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18842         * config/nds32/nds32.c (nds32_option_override): Use
18843         nds32_init_rtx_costs function.
18845 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18847         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18848         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18849         (TARGET_PIPELINE_N8): Likewise.
18850         (TARGET_PIPELINE_N10): Likewise.
18851         (TARGET_PIPELINE_N13): Likewise.
18852         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18854 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18856         * config/nds32/nds32-fpu.md: Update copyright year.
18858 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18860         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18862 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18864         * config/nds32/nds32.c
18865         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18866         * config/nds32/nds32.opt (minline-asm-r15): New option.
18868 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18870         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18871         MASK_HW_ABS.
18872         * config/nds32/nds32.md (abssi2): New pattern.
18874 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
18876         * config/i386/i386.md (rex64namesuffix): New mode attribute.
18877         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18878         Merge insn pattern from sse_cvtsi2ss<round_name> and
18879         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18880         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18881         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18882         using SWI48 mode iterator.
18883         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18884         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18885         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18886         pattern from sse_cvttss2si<round_saeonly_name>
18887         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18888         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18889         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18890         using SWI48 mode iterator.
18891         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18892         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18893         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18894         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18895         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18896         using SWI48 mode iterator.
18897         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18898         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18899         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18900         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18901         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18902         SWI48 mode iterator.
18903         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18904         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18905         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18906         pattern from sse_cvttsd2si<round_saeonly_name>
18907         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18909 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18911         * config/nds32/nds32-md-auxiliary.c
18912         (nds32_valid_smw_lwm_base_p): Refine.
18913         (nds32_output_smw_single_word): Refine.
18914         (nds32_output_smw_double_word): New.
18915         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18917 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18919         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18920         (nds32_output_stack_pop): Refine.
18921         (nds32_expand_unaligned_load): Refine.
18922         (nds32_expand_unaligned_store): Refine.
18924 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18925             Chung-Ju Wu  <jasonwucj@gmail.com>
18927         * config/nds32/constants.md: Add TP_REGNUM constant.
18928         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18929         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18930         UNSPEC_ADD32.
18931         * config/nds32/nds32-doubleword.md: Consider flag_pic.
18932         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18933         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18934         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18935         and PIC code generation.
18936         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18937         code generation.
18938         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18939         optimization.
18940         * config/nds32/nds32.md: Support TLS and PIC.
18941         * config/nds32/nds32.c: Support TLS and PIC.
18942         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18943         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18944         predicate.
18946 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18948         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18949         mode with E_ prefix.
18951 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
18952             Chung-Ju Wu  <jasonwucj@gmail.com>
18954         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18955         * config/nds32/nds32-md-auxiliary.c
18956         (symbolic_reference_mentioned_p): New.
18957         (nds32_legitimize_ict_address): New.
18958         (nds32_expand_ict_move): New.
18959         (nds32_indirect_call_referenced_p): New.
18960         (nds32_symbol_binds_local_p): Delete.
18961         (nds32_long_call_p): Modify.
18962         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18963         * config/nds32/nds32-protos.h
18964         (symbolic_reference_mentioned_p): Declare.
18965         (nds32_legitimize_ict_address): Declare.
18966         (nds32_expand_ict_move): Declare.
18967         (nds32_indirect_call_referenced_p): Declare.
18968         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18969         (nds32_relax_group): Use nds32_ict_const_p as condition.
18970         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18971         (nds32_asm_file_start): Output ict_model directive in asm code.
18972         (nds32_legitimate_address_p): Consider indirect call.
18973         (nds32_print_operand): Consider indirect call.
18974         (nds32_print_operand_address): Consider indirect call.
18975         (nds32_insert_attributes): Handle "indirect_call" attribute.
18976         (TARGET_LEGITIMATE_ADDRESS_P): Define.
18977         (TARGET_LEGITIMATE_CONSTANT_P): Define.
18978         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18979         (TARGET_DELEGITIMIZE_ADDRESS): Define.
18980         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18981         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18982         (TARGET_ICT_MODEL_SMALL): Define.
18983         (TARGET_ICT_MODEL_LARGE): Define.
18984         * config/nds32/nds32.md (movsi): Consider ict model.
18985         (call, call_value): Consider ict model.
18986         (sibcall, sibcall_value): Consider ict model.
18987         * config/nds32/nds32.opt (mict-model): New option.
18988         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18989         model.
18991 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
18992             Monk Chiang  <sh.chiang04@gmail.com>
18993             Jim Wilson <jimw@sifive.com>
18995         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18996         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
18997         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18998         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18999         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19000         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19001         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19002         compute save_libcall_adjustment properly.
19003         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19004         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19005         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19006         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19007         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19008         (ABI_SPEC): Handle mabi=ilp32e.
19009         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19010         (RVE): Add RVE mask.
19011         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19012         <-march>: Add rv32e as an example.
19014 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19016         PR c++/82899
19017         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19018         (intra_create_variable_infos): Handle C++ constructors.
19020 2018-05-18  Martin Liska  <mliska@suse.cz>
19022         * passes.def: Remove a redundant pass.
19024 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19026         PR bootstrap/85838
19027         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19029 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19031         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19032         (ARMv4): Update.
19033         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19034         (ARMv6m): Update.
19035         (armv2, armv2a, armv3, armv3m): Delete architectures.
19036         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19037         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19038         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19039         Delete cpus.
19040         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19041         (*mulsidi3adddi): Likewise.
19042         (mulsidi3): Likewise.
19043         (*mulsidi3_nov6): Likewise.
19044         (umulsidi3): Likewise.
19045         (umulsidi3_nov6): Likewise.
19046         (umaddsidi4): Likewise.
19047         (*umulsidi3adddi): Likewise.
19048         (smulsi3_highpart): Likewise.
19049         (*smulsi3_highpart_nov6): Likewise.
19050         (umulsi3_highpart): Likewise.
19051         (*umulsi3_highpart_nov6): Likewise.
19052         * config/arm/arm.h (arm_arch3m): Delete.
19053         * config/arm/arm.c (arm_arch3m): Delete.
19054         (arm_option_override_internal): Update armv3-related comment.
19055         (arm_configure_build_target): Delete use of isa_bit_mode32.
19056         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19057         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19058         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19059         (mulsa3): Likewise.
19060         (mulusa3): Likewise.
19061         * config/arm/arm-protos.h (arm_arch3m): Delete.
19062         * config/arm/arm-tables.opt: Regenerate.
19063         * config/arm/arm-tune.md: Likewise.
19064         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19065         deleted architectures.
19067 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19069         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19070         (armv5t, armv5te): New features.
19071         (ARMv5, ARMv5e): Delete fgroups.
19072         (ARMv5t, ARMv5te): Adjust for above changes.
19073         (ARMv6m): Likewise.
19074         (armv5, armv5e): Delete arches.
19075         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19076         arm_arch5.
19077         (*call_reg_arm): Likewise.
19078         (*call_value_reg_armv5): Likewise.
19079         (*call_value_reg_arm): Likewise.
19080         (*call_symbol): Likewise.
19081         (*call_value_symbol): Likewise.
19082         (*sibcall_insn): Likewise.
19083         (*sibcall_value_insn): Likewise.
19084         (clzsi2): Likewise.
19085         (prefetch): Likewise.
19086         (define_split and define_peephole2 dependent on arm_arch5):
19087         Likewise.
19088         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19089         arm_arch5e.
19090         (TARGET_ARM_QBIT): Likewise.
19091         (TARGET_DSP_MULTIPLY): Likewise.
19092         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19093         (arm_arch5, arm_arch5e): Delete.
19094         (arm_arch5t, arm_arch5te): Declare.
19095         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19096         (arm_arch5t): Declare.
19097         (arm_option_reconfigure_globals): Update for the above.
19098         (arm_options_perform_arch_sanity_checks): Update comment, replace
19099         use of arm_arch5 with arm_arch5t.
19100         (use_return_insn): Likewise.
19101         (arm_emit_call_insn): Likewise.
19102         (output_return_instruction): Likewise.
19103         (arm_final_prescan_insn): Likewise.
19104         (arm_coproc_builtin_available): Likewise.
19105         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19106         arm_arch5e with arm_arch5t and arm_arch5te.
19107         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19108         (arm_arch5t, arm_arch5te): Declare.
19109         * config/arm/arm-tables.opt: Regenerate.
19110         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19111         * config/arm/t-multilib: Likewise.
19112         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19113         instead of arm_arch5.
19114         (*call_reg_thumb1): Likewise.
19115         (*call_value_reg_thumb1_v5): Likewise.
19116         (*call_value_reg_thumb1): Likewise.
19117         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19118         unreachable path.
19119         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19121 2018-05-18  Martin Liska  <mliska@suse.cz>
19123         PR gcov-profile/84846
19124         * doc/gcov.texi: Document -t option of gcov tool.
19126 2018-05-18  Martin Liska  <mliska@suse.cz>
19128         PR gcov-profile/84846
19129         * gcov.c (print_usage): Add new -t option.
19130         (process_args): Handle the option.
19131         (generate_results): Use stdout as output when requested by
19132         the option.
19134 2018-05-18  Martin Liska  <mliska@suse.cz>
19136         PR gcov-profile/84846
19137         * coverage.c (coverage_init): Write PWD to .gcno file.
19138         * doc/gcov.texi: Document how working directory is printed.
19139         * gcov-dump.c (dump_gcov_file): Print PWD.
19140         * gcov.c (output_intermediate_file): Likewise.
19141         (read_graph_file): Read PWD string.
19142         (output_lines): Print PWD.
19144 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19146         PR middle-end/85817
19147         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19148         for retval and return false if all args to phi are zero.
19150 2018-05-18  Richard Biener  <rguenther@suse.de>
19152         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19153         method.
19154         (evrp_dom_walker::before_dom_children): Call it.
19156 2018-05-18  Richard Biener  <rguenther@suse.de>
19158         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19159         results when processing array refs with variable index.
19161 2018-05-18  Toon Moene  <toon@moene.org>
19163         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19164         directly after that of -floop-interchange. Indicate that both
19165         options are enabled by default when specifying -O3.
19167 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19169         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19170         iterator.  Delete separate integer-mode vec_set<mode> expander.
19171         (aarch64_simd_vec_setv2di): Delete.
19172         (vec_setv2di): Delete.
19173         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19174         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19175         the "w, r" alternative.
19177 2018-05-18  Martin Liska  <mliska@suse.cz>
19179         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19180         * tree-pass.h (make_pass_lower_switch_O0): New function.
19181         * tree-switch-conversion.c (node_has_low_bound): Remove.
19182         (node_has_high_bound): Likewise.
19183         (node_is_bounded): Likewise.
19184         (class pass_lower_switch): Make it a template type and create
19185         two instances.
19186         (pass_lower_switch::execute): Add template argument.
19187         (make_pass_lower_switch): New function.
19188         (make_pass_lower_switch_O0): New function.
19189         (do_jump_if_equal): Remove.
19190         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19191         all the hard work to tree optimization passes.
19193 2018-05-18  Martin Liska  <mliska@suse.cz>
19195         * dbgcnt.c (limit_low): Renamed from limit.
19196         (limit_high): New variable.
19197         (dbg_cnt_is_enabled): Check for upper limit.
19198         (dbg_cnt): Adjust dumping.
19199         (dbg_cnt_set_limit_by_index): Add new argument for high
19200         value.
19201         (dbg_cnt_set_limit_by_name): Likewise.
19202         (dbg_cnt_process_single_pair): Parse new format.
19203         (dbg_cnt_process_opt): Use strtok.
19204         (dbg_cnt_list_all_counters): Remove 'value' and add
19205         'limit_high'.
19206         * doc/invoke.texi: Document changes.
19208 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19210         * doc/sourcebuild.texi (scalar_all_fma): Document.
19211         * tree.def (FMA_EXPR): Delete.
19212         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19213         * internal-fn.c (ternary_direct): New macro.
19214         (expand_ternary_optab_fn): Likewise.
19215         (direct_ternary_optab_supported_p): Likewise.
19216         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19217         * builtins.c (fold_builtin_fma): Delete.
19218         (fold_builtin_3): Don't call it.
19219         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19220         * expr.c (expand_expr_real_2): Likewise.
19221         * fold-const.c (operand_equal_p): Likewise.
19222         (fold_ternary_loc): Likewise.
19223         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19224         * gimple.c (DEFTREECODE): Likewise.
19225         * gimplify.c (gimplify_expr): Likewise.
19226         * optabs-tree.c (optab_for_tree_code): Likewise.
19227         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19228         * tree-eh.c (operation_could_trap_p): Likewise.
19229         (stmt_could_throw_1_p): Likewise.
19230         * tree-inline.c (estimate_operator_cost): Likewise.
19231         * tree-pretty-print.c (dump_generic_node): Likewise.
19232         (op_code_prio): Likewise.
19233         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19234         * tree-ssa-operands.c (get_expr_operands): Likewise.
19235         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19236         * fold-const-call.h (fold_fma): Delete.
19237         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19238         CFN_FNMA and CFN_FNMS.
19239         (fold_fma): Delete.
19240         * genmatch.c (combined_fn): New enum.
19241         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19242         (commutative_op): New function.
19243         (commutate): Use it.  Handle more than 2 operands.
19244         (dt_operand::gen_gimple_expr): Use commutative_op.
19245         (parser::parse_expr): Allow :c to be used with non-binary
19246         operators if the commutative operand is known.
19247         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19248         CFN_FMS, CFN_FNMA and CFN_FNMS.
19249         (backprop::process_assign_use): Remove FMA_EXPR handling.
19250         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19251         (gen_hsa_fma): New function.
19252         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19253         IFN_FNMA and IFN_FNMS.
19254         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19255         * gimple-fold.h (follow_all_ssa_edges): Declare.
19256         * gimple-fold.c (follow_all_ssa_edges): New function.
19257         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19258         gimple_build interface and use follow_all_ssa_edges to fold the result.
19259         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19260         instead of checking for optabs directly.
19261         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19262         rather than FMA_EXPRs.
19263         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19264         call to IFN_FMA instead of an FMA_EXPR.
19266 2018-05-17  Jim Wilson  <jimw@sifive.com>
19268         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19269         sign extended promoted subreg, and the range does not have the sign bit
19270         set, then do a sign extend.
19272         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19273         test, check for sign extended subreg and/or constant operands, and
19274         do a sign extend in that case.
19276 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19278         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19279         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19280         Add untyped.
19281         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19282         Change logics_shift_reg to logics_shift_imm.
19283         (thunderx2t99_fp_loadpair_basic): Delete.
19284         (thunderx2t99_fp_storepair_basic): Delete.
19285         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19286         (thunderx2t99_asimd_polynomial): Delete.
19287         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19288         and neon_fp_mul_d_scalar_q.
19289         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19290         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19291         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19292         (thunderx2t99_asimd_lut): Add missing tbl types.
19293         (thunderx2t99_asimd_ext): Delete.
19294         (thunderx2t99_asimd_load1_1_mult): Delete.
19295         (thunderx2t99_asimd_load1_2_mult): Delete.
19296         (thunderx2t99_asimd_load1_ldp): New.
19297         (thunderx2t99_asimd_load1): New.
19298         (thunderx2t99_asimd_load2): Add missing *load2* types.
19299         (thunderx2t99_asimd_load3): New.
19300         (thunderx2t99_asimd_load4): New.
19301         (thunderx2t99_asimd_store1_1_mult): Delete.
19302         (thunderx2t99_asimd_store1_2_mult): Delete.
19303         (thunderx2t99_asimd_store2_mult): Delete.
19304         (thunderx2t99_asimd_store2_onelane): Delete.
19305         (thunderx2t99_asimd_store_stp): New.
19306         (thunderx2t99_asimd_store1): New.
19307         (thunderx2t99_asimd_store2): New.
19308         (thunderx2t99_asimd_store3): New.
19309         (thunderx2t99_asimd_store4): New.
19311 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19313         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19314         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19316 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19317             Segher Boessenkool  <segher@kernel.crashing.org>
19319         PR target/85698
19320         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19321         operand.
19323 2018-05-17  Richard Biener  <rguenther@suse.de>
19325         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19326         for pruning loop and prune defs feeding only already visited PHIs.
19328 2018-05-17  Richard Biener  <rguenther@suse.de>
19330         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19332 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19333             Richard Biener  <rguenther@suse.de>
19335         PR tree-optimization/85793
19336         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19337         for VMAT_ELEMENTWISE.
19339 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19341         * internal-fn.h (lookup_internal_fn): Declare
19342         * internal-fn.c (lookup_internal_fn): New function.
19343         * gimple.c (gimple_build_call_from_tree): Handle calls to
19344         internal functions.
19345         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19346         internal function names.
19347         * tree-pretty-print.c (dump_generic_node): Likewise.
19348         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19350 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19352         * gimple-fold.h (gimple_build): Make the function forms take
19353         combined_fn rather than built_in_function.
19354         (gimple_simplify): Likewise.
19355         * gimple-match-head.c (gimple_simplify): Likewise.
19356         * gimple-fold.c (gimple_build): Likewise.
19357         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19358         rather than gimple_build_call_internal.
19359         (get_initial_defs_for_reduction): Likewise.
19360         (vect_create_epilog_for_reduction): Likewise.
19361         (vectorizable_live_operation): Likewise.
19363 2018-05-17  Martin Liska  <mliska@suse.cz>
19365         * gimple-ssa-sprintf.c (format_directive): Do not use
19366         space in between 'G_' and '('.
19368 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19370         PR target/85323
19371         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19372         even if the mask is not all ones.
19374         PR target/85323
19375         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19376         vector.
19377         (ix86_gimple_fold_builtin): Likewise.
19379         PR target/85323
19380         * config/i386/i386.c: Include tree-vector-builder.h.
19381         (ix86_vector_shift_count): New function.
19382         (ix86_fold_builtin): Fold shift builtins by scalar count.
19383         (ix86_gimple_fold_builtin): Likewise.
19385         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19386         _mm512_setzero): New intrinsics.
19388 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19389             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19391         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19392         code generation for cases where splatting a value is not useful.
19393         * simplify-rtx.c (simplify_ternary_operation): Simplify
19394         vec_merge across a vec_duplicate and a paradoxical subreg forming
19395         a vector mode to a vec_concat.
19397 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19399         * config.gcc: Support "goldmont-plus".
19400         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19401         "goldmont-plus".
19402         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19403         PROCESSOR_GOLDMONT_PLUS.
19404         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19405         (processor_target_table): Add "goldmont-plus".
19406         (PTA_GOLDMONT_PLUS): Define.
19407         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19408         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19409         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19410         (fold_builtin_cpu): Add "goldmont-plus".
19411         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19412         (ix86_option_override_internal): Add "goldmont-plus".
19413         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19414         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19415         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19416         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19418 2018-05-17  Richard Biener  <rguenther@suse.de>
19420         PR tree-optimization/85757
19421         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19422         remove defs that only feed that PHI from further processing.
19424 2018-05-16  Jim Wilson  <jimw@sifive.com>
19426         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19427         asterisk to name.
19428         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19430 2018-05-16  Mark Wielaard  <mark@klomp.org>
19432         * dwarf2out.c (count_index_strings): New function.
19433         (output_indirect_strings): Call count_index_strings and generate
19434         header for dwarf_version >= 5.
19436 2018-05-16  Mark Wielaard  <mark@klomp.org>
19438         * dwarf2out.c (dwarf_FORM): New function.
19439         (set_indirect_string): Use dwarf_FORM.
19440         (reset_indirect_string): Likewise.
19441         (size_of_die): Likewise.
19442         (value_format): Likewise.
19443         (output_die): Likewise.
19444         (add_skeleton_AT_string): Likewise.
19445         (output_macinfo_op): Likewise.
19446         (index_string): Likewise.
19447         (output_index_string_offset): Likewise.
19448         (output_index_string): Likewise.
19449         (count_index_strings): Likewise.
19451 2018-05-16  Carl Love  <cel@us.ibm.com>
19453         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19454         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19456 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19458         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19459         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19461 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19463         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19464         (fnma<mode>4): Likewise.
19465         (fms<mode>4): Likewise.
19466         (fnms<mode>4): Likewise.
19467         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19468         (aarch64_fnma<mode>4): Likewise.
19469         (aarch64_fms<mode>4): Likewise.
19470         (aarch64_fnms<mode>4): Likewise.
19471         (aarch64_fnmadd<mode>4): Likewise.
19473 2018-05-16  Jason Merrill  <jason@redhat.com>
19475         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19477 2018-05-16  Richard Biener  <rguenther@suse.de>
19479         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19480         (dump_stmt_cost): Declare.
19481         (add_stmt_cost): Dump cost we add.
19482         (add_stmt_costs): New function.
19483         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19484         No longer exported.
19485         (vect_analyze_stmt): Adjust prototype.
19486         (vectorizable_condition): Likewise.
19487         (vectorizable_live_operation): Likewise.
19488         (vectorizable_reduction): Likewise.
19489         (vectorizable_induction): Likewise.
19490         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19491         cost vector to pass to vectorizable_ and record afterwards.
19492         (vect_model_reduction_cost): Take cost vector argument and adjust.
19493         (vect_model_induction_cost): Likewise.
19494         (vectorizable_reduction): Likewise.
19495         (vectorizable_induction): Likewise.
19496         (vectorizable_live_operation): Likewise.
19497         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19498         SLP_TREE_NUMBER_OF_VEC_STMTS.
19499         (vect_analyze_slp_cost_1): Remove.
19500         (vect_analyze_slp_cost): Likewise.
19501         (vect_slp_analyze_node_operations): Take visited args and
19502         a target cost vector.  Avoid processing already visited stmt sets.
19503         (vect_slp_analyze_operations): Use a local cost vector to gather
19504         costs and register those of non-discarded instances.
19505         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19506         (vect_schedule_slp_instance): Remove copying of
19507         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19508         zero.
19509         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19510         adding cost.  Record cost entry location.
19511         (vect_prologue_cost_for_slp_op): Function to compute cost of
19512         a constant or invariant generated for SLP vect in the prologue,
19513         split out from vect_analyze_slp_cost_1.
19514         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19515         (vect_model_promotion_demotion_cost): Likewise.
19516         (vect_model_store_cost): Likewise, make static.
19517         (vect_model_load_cost): Likewise.
19518         (vectorizable_bswap): Add cost vector arg and adjust.
19519         (vectorizable_call): Likewise.
19520         (vectorizable_simd_clone_call): Likewise.
19521         (vectorizable_conversion): Likewise.
19522         (vectorizable_assignment): Likewise.
19523         (vectorizable_shift): Likewise.
19524         (vectorizable_operation): Likewise.
19525         (vectorizable_store): Likewise.
19526         (vectorizable_load): Likewise.
19527         (vectorizable_condition): Likewise.
19528         (vectorizable_comparison): Likewise.
19529         (can_vectorize_live_stmts): Likewise.
19530         (vect_analyze_stmt): Likewise.
19531         (vect_transform_stmt): Adjust calls to vectorizable_*.
19532         * tree-vectorizer.c: Include gimple-pretty-print.h.
19533         (dump_stmt_cost): New function.
19535 2018-05-16  Richard Biener  <rguenther@suse.de>
19537         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19538         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19539         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19540         (check_name): New callback.
19541         (dse_classify_store): Track cycles via a visited bitmap of PHI
19542         defs and simplify handling of in-loop and across loop dead stores
19543         and properly fail for loop-variant refs.  Handle byte-tracking with
19544         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19545         limiting the walk.
19547 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19549         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19550         (vect_get_mask_type_for_stmt): Likewise.
19551         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19552         split out from...
19553         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19554         to determine the statement's vector type and the vector type that
19555         should be used for calculating nunits.  Deal with cases in which
19556         the type has to be deferred.
19557         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19558         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19559         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19560         (vect_determine_vf_for_stmt): New functions, split out from...
19561         (vect_determine_vectorization_factor): ...here.
19562         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19563         (vect_get_mask_type_for_stmt): New functions, split out from
19564         vect_determine_vectorization_factor.
19566 2018-05-16  Richard Biener  <rguenther@suse.de>
19568         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19569         verify the [VEC_]COND_EXPR embedded comparison.
19571 2018-05-15  Martin Sebor  <msebor@redhat.com>
19573         PR tree-optimization/85753
19574         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19575         RECORD_TYPE in addition to ARRAY_TYPE.
19577 2018-05-15  Martin Sebor  <msebor@redhat.com>
19579         PR middle-end/85643
19580         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19582 2018-05-15  Richard Biener  <rguenther@suse.de>
19584         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19585         add by_clobber_p one.  Change algorithm to collect all defs
19586         representing uses we need to walk and try reducing them to
19587         a single one before failing.
19588         (dse_dom_walker::dse_optimize_stmt): Adjust.
19590 2018-05-13  Mark Wielaard  <mark@klomp.org>
19592         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19593         (size_of_loc_descr): Likewise.
19594         (output_loc_operands): Likewise.
19595         (output_loc_operands_raw): Likewise.
19596         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19597         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19598         (hash_loc_operands): Likewise.
19599         (compare_loc_operands): Likewise.
19601 2018-05-14  Mark Wielaard  <mark@klomp.org>
19603         * dwarf2out.c (count_index_addrs): New function.
19604         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19606 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19608         PR tree-optimization/83648
19609         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19610         return value as malloc candidate.
19612 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19614         PR ipa/85734
19615         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19616         param as true in call to suggest_attribute.
19618 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19620         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19621         -mreadonly-in-sdata.
19623 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19625         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19626         New pattern.
19627         (aarch64_crypto_aesd_fused): Likewise.
19629 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19631         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19632         (movsi_aarch64): Likewise.
19633         (load_pairsi): Likewise.
19634         (load_pairdi): Likewise.
19635         (store_pairsi): Likewise.
19636         (store_pairdi): Likewise.
19637         (load_pairsf): Likewise.
19638         (load_pairdf): Likewise.
19639         (store_pairsf): Likewise.
19640         (store_pairdf): Likewise.
19641         (zero_extend): Likewise.
19642         (trunc): Swap alternatives.
19643         (fcvt_target): Add '?' to prefer w over r.
19645 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19647         PR target/85756
19648         * config/i386/i386.md: Disallow non-commutative arithmetics in
19649         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19650         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19651         in the peephole2 before it.
19653 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19655         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19656         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19657         (ix86_handle_option): Handle -mcldemote.
19658         * config.gcc: New header.
19659         * config/i386/cldemoteintrin.h: New file.
19660         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19661         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19662         -mcldemote.
19663         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19664         OPTION_MASK_ISA_CLDEMOTE.
19665         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19666         (ix86_valid_target_attribute_inner_p): Ditto.
19667         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19668         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19669         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19670         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19671         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19672         (cldemote): New.
19673         * config/i386/i386.opt: Add -mcldemote.
19674         * config/i386/x86intrin.h: New header.
19675         * doc/invoke.texi: Add -mcldemote.
19677 2018-05-14  Richard Biener  <rguenther@suse.de>
19679         * doc/match-and-simplify.texi: Adjust :s documentation.
19681 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19683         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19684         intended memcpy size.
19685         (REORDER_45): Likewise.
19687 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19689         * sort.cc: New file.
19690         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19691         * vec.c (qsort_chk): Use gcc_qsort.
19692         * Makefile.in (OBJS-libcommon): Add sort.o.
19693         (build/sort.o): New target.  Use it...
19694         (BUILD_RTL): ... here, and...
19695         (build/gencfn-macros): ... here, and...
19696         (build/genmatch): ... here.
19698 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19699             Chung-Ju Wu  <jasonwucj@gmail.com>
19701         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19702         * config/nds32/nds32-graywolf.md: New file.
19703         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19704         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19705         pipeline.
19706         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19707         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19708         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19709         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19710         * config/nds32/pipelines.md: Include n15 settings.
19712 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19713             Chung-Ju Wu  <jasonwucj@gmail.com>
19715         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19716         * config/nds32/nds32-n13.md: New file.
19717         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19718         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19719         pipeline.
19720         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19721         * config/nds32/nds32.md (pipeline_model): Add n13.
19722         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19723         * config/nds32/pipelines.md: Include n13 settings.
19725 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19726             Chung-Ju Wu  <jasonwucj@gmail.com>
19728         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19729         * config/nds32/nds32-n10.md: New file.
19730         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19731         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19732         pipeline.
19733         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19734         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19735         * config/nds32/nds32.md (pipeline_model): Add n10.
19736         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19737         * config/nds32/pipelines.md: Include n10 settings.
19739 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19740             Kito Cheng  <kito.cheng@gmail.com>
19741             Chung-Ju Wu  <jasonwucj@gmail.com>
19743         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19744         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19745         Add enum values for DSP extension instructions.
19746         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19747         New constraints.
19748         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19749         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19750         New code iterators.
19751         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19752         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19753         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19754         * config/nds32/nds32-intrinsic.md: Likewise.
19755         * config/nds32/nds32_intrinsic.h: Likewise.
19756         * config/nds32/nds32-md-auxiliary.c: Likewise.
19757         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19758         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19759         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19760         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19761         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19762         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19763         TYPE_DMAC in switch statement.
19764         * config/nds32/nds32.c: New checking and implementation for DSP
19765         extension instructions.
19766         * config/nds32/nds32.h: Likewise.
19767         * config/nds32/nds32.md: Likewise.
19768         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19769         * config/nds32/predicates.md: Implement new predicates for DSP
19770         extension.
19772 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19774         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19775         Reformat alternatives and attributes so it is easier to identify
19776         which constraints/attributes go with which instruction.
19777         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19778         (mov<mode>_softfloat32, FMOVE64): Likewise.
19779         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19780         (mov<mode>_softfloat64, FMOVE64): Likewise.
19782 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19784         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19785         subsection.
19786         (Basic PowerPC Built-in Functions): The new name of the
19787         subsection previously known as "PowerPC Built-in Functions".
19788         (Basic PowerPC Built-in Functions Available on all Configurations):
19789         New subsubsection.
19790         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19791         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19792         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19793         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19795 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19797         PR ipa/85655
19798         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19799         single const.
19801 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19803         PR target/85733
19804         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19806 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19808         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19809         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19810         (ix86_handle_option): Handle -mwaitpkg.
19811         * config.gcc: New header.
19812         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19813         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19814         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19815         function type.
19816         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19817         OPTION_MASK_ISA_WAITPKG.
19818         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19819         (ix86_option_override_internal): Add PTA_WAITPKG.
19820         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19821         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19822         IX86_BUILTIN_TPAUSE.
19823         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19824         __builtin_ia32_umwait and __builtin_ia32_tpause.
19825         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19826         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19827         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19828         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19829         UNSPECV_TPAUSE): New.
19830         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19831         * config/i386/i386.opt: Add -mwaitpkg.
19832         * config/i386/waitpkgintrin.h: New file.
19833         * config/i386/x86intrin.h: New header.
19834         * doc/invoke.texi: Add -mwaitpkg.
19836 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19838         PR target/85606
19839         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19840         equivalent.
19841         (cortex-m0): Use armv6s-m isa.
19842         (cortex-m0plus): Likewise.
19843         (cortex-m1): Likewise.
19844         (cortex-m0.small-multiply): Likewise.
19845         (cortex-m0plus.small-multiply): Likewise.
19846         (cortex-m1.small-multiply): Likewise.
19848 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19849             Jakub Jelinek  <jakub@redhat.com>
19851         PR tree-optimization/85692
19852         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19853         source permute as well.
19855 2018-05-11  Martin Liska  <mliska@suse.cz>
19857         PR sanitizer/85556
19858         * doc/extend.texi: Document LLVM style format for no_sanitize
19859         attribute.
19861 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19863         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19864         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19865         (mode_supports_vsx_dform_quad): Likewise.
19866         (mode_supports_vmx_dform): Move these functions to be next to the
19867         other mode_supports functions.
19868         (mode_supports_dq_form): Likewise.
19869         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19870         mode_supports_dq_form.
19871         (reg_offset_addressing_ok_p): Likewise.
19872         (offsettable_ok_by_alignment): Likewise.
19873         (rs6000_legitimate_offset_address_p): Likewise.
19874         (legitimate_lo_sum_address_p): Likewise.
19875         (rs6000_legitimize_address): Likewise.
19876         (rs6000_legitimize_reload_address): Likewise.
19877         (rs6000_secondary_reload_inner): Likewise.
19878         (rs6000_preferred_reload_class): Likewise.
19879         (rs6000_output_move_128bit): Likewise.
19881 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19883         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19884         Generate SImode target register for null target.
19885         <case IX86_BUILTIN_XGETBV>: Ditto.
19886         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19887         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19889 2018-05-10  Carl Love  <cel@us.ibm.com>
19891         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19892         dcbtt and dcbtstt if operands[2] is 0.
19894 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
19896         PR target/85693
19897         * config/i386/sse.md (usadv64qi): New expander.
19899 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
19901         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19902         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19903         -maltivec=be support.
19904         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19905         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19906         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19907         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19908         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19909         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19910         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19911         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19912         altivec_vsumsws): Adjust.
19913         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19914         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19915         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19916         support.
19917         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19918         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19919         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19920         (altivec_lve<VI_char>x): Delete expand.
19921         (*altivec_lve<VI_char>x_internal): Rename to...
19922         (altivec_lve<VI_char>x): ... this.
19923         (altivec_lvxl_<mode>): Delete expand.
19924         (*altivec_lvxl_<mode>_internal): Rename to ...
19925         (altivec_lvxl_<mode>): ... this.
19926         (altivec_stvxl_<mode>): Delete expand.
19927         (*altivec_stvxl_<mode>_internal): Rename to ...
19928         (altivec_stvxl_<mode>): ... this.
19929         (altivec_stve<VI_char>x): Delete expand.
19930         (*altivec_stve<VI_char>x_internal): Rename to ...
19931         (altivec_stve<VI_char>x): ... this.
19932         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19933         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19934         reduc_plus_scal_<mode>): Adjust.
19935         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19936         comment.
19937         (rs6000_cpu_cpp_builtins): Adjust.
19938         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19939         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19940         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19941         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19942         -maltivec=be support.
19943         (rs6000_split_vec_extract_var): Adjust.
19944         (rs6000_split_v4si_init): Adjust.
19945         (swap_selector_for_mode): Delete.
19946         (altivec_expand_lvx_be, altivec_expand_stvx_be,
19947         altivec_expand_stvex_be): Delete.
19948         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19949         -maltivec=be support.
19950         (rs6000_gimple_fold_builtin): Ditto.
19951         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19952         Adjust.
19953         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19954         (TARGET_DIRECT_MOVE_64BIT): Adjust.
19955         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19956         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19957         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19958         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19959         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19960         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19961         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19962         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19963         anonymous split): Adjust.
19964         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19965         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19967 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
19969         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19970         when --with-gxx-include-dir is also specified.
19971         * configure: Regenerate.
19973 2018-05-09  Jim Wilson  <jimw@sifive.com>
19975         PR target/84797
19976         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19977         * config/riscv/t-withmultilib: New.
19978         * config/riscv/withmultilib.h: New.
19979         * doc/install.texi: Document RISC-V --with-multilib-list support.
19981 2018-05-09  Richard Biener  <rguenther@suse.de>
19983         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19984         vector.
19985         (vect_bb_vectorization_profitable_p): Adjust.  Compute
19986         actual scalar cost using the cost vector and the add_stmt_cost
19987         machinery.
19989 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19991         PR rtl-optimization/85645
19992         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19993         in the REG_CFA_REGISTER note for LR, don't leave it empty.
19995 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
19997         PR rtl-optimization/85645
19998         * shrink-wrap.c (spread_components): Return a boolean saying if
19999         anything was changed.
20000         (try_shrink_wrapping_separate): Iterate spread_components until
20001         nothing changes anymore.
20003 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20005         PR rtl-optimization/85645
20006         * regrename.c (build_def_use): Also kill the chains that include the
20007         destination of a REG_CFA_REGISTER note.
20009 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20011         PR rtl-optimization/85645
20012         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20013         insn that has a REG_CFA_REGISTER note.
20015 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20017         * cfgexpand.c (expand_clobber): New function.
20018         (expand_gimple_stmt_1): Use it.
20019         * tree-vect-stmts.c (vect_clobber_variable): New function,
20020         split out from...
20021         (vectorizable_simd_clone_call): ...here.
20022         (vectorizable_store): Emit a clobber either side of an
20023         IFN_STORE_LANES sequence.
20024         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20026 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20028         PR target/85626
20029         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20030         (define_insn "trap_if_false"): Add exit after trap.
20032 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20034         PR rtl-optimization/85638
20035         * bb-reorder.c: Include common/common-target.h.
20036         (create_forwarder_block): New function extracted from...
20037         (fix_up_crossing_landing_pad): ...here.  Rename into...
20038         (dw2_fix_up_crossing_landing_pad): ...this.
20039         (sjlj_fix_up_crossing_landing_pad): New function.
20040         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20041         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20042         from both partitions and exit the loop after one iteration.
20044 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20046         Revert:
20047         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20048         subsection.
20049         (Basic PowerPC Built-in Functions): The new name of the
20050         subsection previously known as "PowerPC Built-in Functions".
20051         (Basic PowerPC Built-in Functions Available on all Configurations):
20052         New subsubsection.
20053         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20054         subsubsection.
20055         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20056         subsubsection.
20057         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20058         subsubsection.
20059         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20060         subsubsection.
20062 2018-05-08  Jim Wilson  <jimw@sifive.com>
20064         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20065         (LD_EMUL_SUFFIX): New.
20066         (LINK_SPEC): Use it.
20068 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20070         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20071         subsection.
20072         (Basic PowerPC Built-in Functions): The new name of the
20073         subsection previously known as "PowerPC Built-in Functions".
20074         (Basic PowerPC Built-in Functions Available on all Configurations):
20075         New subsubsection.
20076         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20077         subsubsection.
20078         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20079         subsubsection.
20080         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20081         subsubsection.
20082         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20083         subsubsection.
20085 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20087         PR target/85683
20088         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20089         after cmpelim optimization.
20091 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20093         * config.gcc: Support "goldmont".
20094         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20095         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20096         PROCESSOR_GOLDMONT.
20097         * config/i386/i386.c (m_GOLDMONT): Define.
20098         (processor_target_table): Add "goldmont".
20099         (PTA_GOLDMONT): Define.
20100         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20101         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20102         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20103         (fold_builtin_cpu): Add "goldmont".
20104         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20105         (ix86_option_override_internal): Add "goldmont".
20106         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20107         (processor_type): Add PROCESSOR_GOLDMONT.
20108         * config/i386/i386.md: Add CPU "glm".
20109         * config/i386/glm.md: New file.
20110         * config/i386/x86-tune.def: Add m_GOLDMONT.
20111         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20113 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20115         PR target/85572
20116         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20117         E_V4DImode.
20118         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20119         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20120         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20122         PR target/85317
20123         * config/i386/i386.c (ix86_fold_builtin): Handle
20124         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20126         PR target/85480
20127         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20128         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20130 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20132         PR target/85658
20133         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20134         (check_arch): Likewise.
20135         (check_fpu): Return the result rather than printing it.
20136         (end arch): Fix operator precedence.
20137         (end cpu): Likewise.
20138         (END): Print the result from check_fpu.
20140 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20141             Alan Hayward  <alan.hayward@arm.com>
20142             David Sherwood  <david.sherwood@arm.com>
20144         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20145         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20146         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20147         (*fcmuo<mode>_and): New patterns.
20149 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20151         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20152         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20153         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20154         (cmp_op, sve_imm_con): New code attributes.
20155         (SVE_COND_INT_CMP, imm_con): Delete.
20156         (cmp_op): Remove above unspecs from int attribute.
20157         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20158         to...
20159         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20160         comparison-specific unspecs.
20161         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20162         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20163         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20164         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20165         (*vec_fcm<cmp_op><mode>): Rename to...
20166         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20167         (*vec_fcmuo<mode>): Rename to...
20168         (*fcmuo<mode>): ...this and adjust likewise.
20169         (*pred_fcm<cmp_op><mode>): New pattern.
20170         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20171         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20172         functions.
20173         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20174         and UNORDERED.
20175         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20176         (aarch64_emit_sve_predicated_cond): New function.
20177         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20178         (aarch64_emit_unspec_cond_or): Replace with...
20179         (aarch64_emit_sve_or_conds): ...this new function.  Use
20180         aarch64_emit_sve_ptrue_op for the individual comparisons and
20181         aarch64_emit_binop to OR them together.
20182         (aarch64_emit_inverted_unspec_cond): Replace with...
20183         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20184         aarch64_emit_sve_ptrue_op for the comparison and
20185         aarch64_emit_unop to invert the result.
20186         (aarch64_expand_sve_vec_cmp_float): Update after the above
20187         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20189 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20191         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20192         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20193         (Backwards Compatibility): Likewise.
20195 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20197         PR bootstrap/85681
20198         Revert:
20199         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20201         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20202         <prefetch_dynamic_strides>: New const bool field.
20203         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20204         prefetch_dynamic_strides.
20205         (exynosm1_prefetch_tune): Likewise.
20206         (thunderxt88_prefetch_tune): Likewise.
20207         (thunderx_prefetch_tune): Likewise.
20208         (thunderx2t99_prefetch_tune): Likewise.
20209         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20210         to false.
20211         (aarch64_override_options_internal): Update to set
20212         PARAM_PREFETCH_DYNAMIC_STRIDES.
20213         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20214         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20215         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20216         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20217         prefetch-dynamic-strides setting.
20219         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20221         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20222         <minimum_stride>: New const int field.
20223         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20224         minimum_stride field.
20225         (exynosm1_prefetch_tune): Likewise.
20226         (thunderxt88_prefetch_tune): Likewise.
20227         (thunderx_prefetch_tune): Likewise.
20228         (thunderx2t99_prefetch_tune): Likewise.
20229         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20230         (aarch64_override_options_internal): Update to set
20231         PARAM_PREFETCH_MINIMUM_STRIDE.
20232         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20233         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20234         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20235         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20236         stride is constant and is below the minimum stride threshold.
20238 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20240         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20241         to 512.
20243 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20245         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20246         <prefetch_dynamic_strides>: New const bool field.
20247         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20248         prefetch_dynamic_strides.
20249         (exynosm1_prefetch_tune): Likewise.
20250         (thunderxt88_prefetch_tune): Likewise.
20251         (thunderx_prefetch_tune): Likewise.
20252         (thunderx2t99_prefetch_tune): Likewise.
20253         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20254         to false.
20255         (aarch64_override_options_internal): Update to set
20256         PARAM_PREFETCH_DYNAMIC_STRIDES.
20257         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20258         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20259         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20260         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20261         prefetch-dynamic-strides setting.
20263 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20265         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20266         <minimum_stride>: New const int field.
20267         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20268         minimum_stride field.
20269         (exynosm1_prefetch_tune): Likewise.
20270         (thunderxt88_prefetch_tune): Likewise.
20271         (thunderx_prefetch_tune): Likewise.
20272         (thunderx2t99_prefetch_tune): Likewise.
20273         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20274         (aarch64_override_options_internal): Update to set
20275         PARAM_PREFETCH_MINIMUM_STRIDE.
20276         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20277         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20278         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20279         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20280         stride is constant and is below the minimum stride threshold.
20282 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20284         PR c++/85659
20285         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20286         the type is addressable.  Don't force op into register if it has
20287         BLKmode.
20289 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20291         PR other/77609
20292         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20293         any section for which we don't know a specific type it should have,
20294         regardless of name.  Previously this was done only for the exact
20295         names ".init_array", ".fini_array", and ".preinit_array".
20296         (default_elf_asm_named_section): Add comment about
20297         relationship with default_section_type_flags and SECTION_NOTYPE.
20298         (get_section): Don't consider it a type conflict if one side has
20299         SECTION_NOTYPE and the other doesn't, as long as neither has the
20300         SECTION_BSS et al used in the default_section_type_flags logic.
20302 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20304         PR target/85653
20305         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20306         (workaround_barsyncs): New function.
20307         (nvptx_reorg): Use workaround_barsyncs.
20308         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20309         (define_expand "nvptx_membar_cta"): New define_expand.
20310         (define_insn "*nvptx_membar_cta"): New insn.
20312 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20314         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20315         To improve optimization opportunities.
20316         * builtin-types.def: The new needed builtin types for the above.
20318 2018-05-04  Richard Biener  <rguenther@suse.de>
20320         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20321         * gimple-ssa-store-merging.c
20322         (imm_store_chain_info::output_merged_store): Remove redundant create,
20323         release split_store vector contents on failure.
20324         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20325         scalar stmt vector on cache hit.
20327 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20329         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20330         Xilinx FP support.
20331         * config.gcc (powerpc-xilinx-eabi*): Remove.
20332         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20333         support.
20334         (fusion_addis_mem_combo_load): Ditto.
20335         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20336         FP support.
20337         (rs6000_cpu_cpp_builtins): Ditto.
20338         * config/rs6000/rs6000-linux.c
20339         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20340         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20341         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20342         support.
20343         (rs6000_setup_reg_addr_masks): Ditto.
20344         (rs6000_init_hard_regno_mode_ok): Ditto.
20345         (rs6000_option_override_internal): Ditto.
20346         (legitimate_lo_sum_address_p): Ditto.
20347         (rs6000_legitimize_address): Ditto.
20348         (rs6000_legitimize_reload_address): Ditto.
20349         (rs6000_legitimate_address_p): Ditto.
20350         (abi_v4_pass_in_fpr): Ditto.
20351         (setup_incoming_varargs): Ditto.
20352         (rs6000_gimplify_va_arg): Ditto.
20353         (rs6000_split_multireg_move): Ditto.
20354         (rs6000_savres_strategy): Ditto.
20355         (rs6000_emit_prologue_components): Ditto.
20356         (rs6000_emit_epilogue_components): Ditto.
20357         (rs6000_emit_prologue): Ditto.
20358         (rs6000_emit_epilogue): Ditto.
20359         (rs6000_elf_file_end): Ditto.
20360         (rs6000_function_value): Ditto.
20361         (rs6000_libcall_value): Ditto.
20362         * config/rs6000/rs6000.h: Ditto.
20363         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20364         (TARGET_MINMAX): ... this.  New.
20365         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20366         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20367         (*movsi_internal1_single): Delete.
20368         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20369         mfpu=, mxilinx-fpu): Delete.
20370         * config/rs6000/singlefp.h: Delete.
20371         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20372         * config/rs6000/t-rs6000: Ditto.
20373         * config/rs6000/t-xilinx: Delete.
20374         * config/rs6000/titan.md: Adjust for fp_type removal.
20375         * config/rs6000/vsx.md: Remove Xilinx FP support.
20376         (VStype_simple): Delete.
20377         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20378         * config/rs6000/xfpu.h: Delete.
20379         * config/rs6000/xfpu.md: Delete.
20380         * config/rs6000/xilinx.h: Delete.
20381         * config/rs6000/xilinx.opt: Delete.
20382         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20383         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20385 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20387         PR libgomp/85639
20388         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20389         if ignore == 0.
20391 2018-05-04  Richard Biener  <rguenther@suse.de>
20393         PR middle-end/85627
20394         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20395         (expand_complex_div_wide): Likewise.
20396         (expand_complex_operations_1): Likewise.
20397         (expand_complex_libcall): Preserve EH info of the original stmt.
20398         (tree_lower_complex): Handle removed blocks.
20399         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20400         on complex multiplication and division libcall builtins.
20402 2018-05-04  Richard Biener  <rguenther@suse.de>
20404         PR middle-end/85574
20405         * fold-const.c (negate_expr_p): Restrict negation of operand
20406         zero of a division to when we know that can happen without
20407         overflow.
20408         (fold_negate_expr_1): Likewise.
20410 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20412         PR libstdc++/85466
20413         * real.h (real_nextafter): Declare.
20414         * real.c (real_nextafter): New function.
20415         * fold-const-call.c (fold_const_nextafter): New function.
20416         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20417         CASE_CFN_NEXTTOWARD.
20418         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20419         even when arg1_mode is different from arg0_mode.
20421 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20423         * doc/extend.texi (Deprecated Features): Remove
20424         -ffriend-injection.
20425         (Backwards Compatibility): Likewise.
20426         * doc/invoke.texi (C++ Language Options): Likewise.
20427         (C++ Dialect Options): Likewise.
20429 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20431         PR target/85530
20432         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20433         _mm512_mask_mullox_epi64): New intrinsics.
20435 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20437         PR testsuite/85106
20438         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20439         dump files): Add offload-tree.
20441 2018-05-03  Richard Biener  <rguenther@suse.de>
20443         PR tree-optimization/85615
20444         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20445         to loops not nested in BBs loop father to avoid creating multi-entry
20446         loops.
20448 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20450         PR tree-optimization/70291
20451         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20452         arguments.  Change return type to tree.  Emit libcall as a new
20453         statement rather than replacing existing one when inplace_p is true.
20454         (expand_complex_multiplication_components): New function.
20455         (expand_complex_multiplication): Expand floating-point complex
20456         multiplication using the above.
20457         (expand_complex_division): Rename inner_type parameter to type.
20458         Update expand_complex_libcall call-site.
20459         (expand_complex_operations_1): Update expand_complex_multiplication
20460         and expand_complex_division call-sites.
20462 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20464         PR target/85582
20465         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20466         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20467         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20468         the highest significant bit of the shift count mask is clear.  In
20469         check whether and[sq]i3 is needed verify that all significant bits
20470         of the shift count other than the highest are set.
20472 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20474         PR libgomp/82428
20475         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20476         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20477         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20478         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20479         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20480         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20481         __builtin_goacc_parlevel_size.
20483 2018-05-02  Richard Biener  <rguenther@suse.de>
20485         PR tree-optimization/85597
20486         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20487         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20488         directly.
20490 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20492         PR testsuite/85106
20493         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20494         dump files): Add ltrans-tree.
20496 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20498         PR testsuite/85106
20499         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20500         dump files): Add wpa-ipa.
20502 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20504         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20505         powerpc*-*-linux*paired* target.
20506         * config/rs6000/750cl.h: Delete.
20507         * config/rs6000/paired.h: Delete.
20508         * config/rs6000/paired.md: Delete.
20509         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20510         float support.
20511         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20512         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20513         comment.  Remove paired float support.
20514         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20515         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20516         VECTOR_PAIRED.
20517         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20518         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20519         declarations.
20520         * config/rs6000/rs6000.c: Remove paired float support.
20521         (paired_expand_vector_init, paired_expand_vector_move,
20522         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20523         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20524         paired_expand_builtin, paired_expand_predicate_builtin,
20525         paired_init_builtins): Delete.
20526         * config/rs6000/rs6000.h: Remove paired float support.
20527         * config/rs6000/rs6000.md: Remove paired float support.
20528         (move_from_CR_ov_bit): Delete.
20529         * config/rs6000/rs6000.opt (mpaired): Delete.
20530         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20531         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20533 2018-05-02  Richard Biener  <rguenther@suse.de>
20535         PR middle-end/85567
20536         * gimplify.c (gimplify_save_expr): When in SSA form allow
20537         SAVE_EXPRs to compute to SSA vars.
20539 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20541         PR target/85582
20542         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20543         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20544         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20545         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20547 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20549         PR tree-optimization/85586
20550         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20551         exit early for statements in the same group if the accesses are
20552         not strided.
20554 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20556         PR lto/85451
20557         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20558         error message.
20560 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20562         PR tree-optimization/85143
20563         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20565 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20567         PR lto/85451
20568         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20569         not found" error message.
20571 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20573         PR other/83786
20574         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20575         * vec.c (test_ordered_remove_if): New function.
20576         (vec_c_tests): Call test_ordered_remove_if.
20577         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20578         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20579         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20580         VEC_ORDERED_REMOVE_IF.
20582 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20584         PR tree-optimization/82665
20585         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20586         pointer subtraction where arguments come from a memchr call.
20588 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20590         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20591         --push-state --as-needed and --pop-state instead of --as-needed and
20592         --no-as-needed if ld supports it.
20593         * configure: Regenerated.
20595         PR web/85578
20596         * doc/install.texi2html: Replace _002d with - and _002a with * in
20597         generated html files using sed.
20599 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20601         PR c++/85523
20602         * gcc-rich-location.c (blank_line_before_p): New function.
20603         (use_new_line): New function.
20604         (gcc_rich_location::add_fixit_insert_formatted): New function.
20605         * gcc-rich-location.h
20606         (gcc_rich_location::add_fixit_insert_formatted): New function.
20608 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20610         * selftest.c (assert_streq): Rename "expected" and "actual" to
20611         "val1" and "val2".  Extend NULL-handling to cover both inputs
20612         symmetrically, while still requiring both to be non-NULL for a pass.
20613         * selftest.h (assert_streq): Rename "expected" and "actual" to
20614         "val1" and "val2".
20615         (ASSERT_EQ): Likewise.
20616         (ASSERT_EQ_AT): Likewise.
20617         (ASSERT_KNOWN_EQ): Likewise.
20618         (ASSERT_KNOWN_EQ_AT): Likewise.
20619         (ASSERT_NE): Likewise.
20620         (ASSERT_MAYBE_NE): Likewise.
20621         (ASSERT_MAYBE_NE_AT): Likewise.
20622         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20623         the assertion to pass.
20624         (ASSERT_STREQ_AT): Likewise.
20626 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20628         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20629         interaction with -pie.
20631 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20633         * selftest.h: Fix alphabetization of per-source-file selftest
20634         declarations.
20636 2018-04-30  Jason Merrill  <jason@redhat.com>
20638         PR c++/61982 - dead stores to destroyed objects.
20639         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20640         of clobber.
20642 2018-04-30  Jason Merrill  <jason@redhat.com>
20644         * tree.c (build_clobber): New.
20645         * tree.h: Declare it.
20646         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20648 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20650         * diagnostic-show-locus.c (layout::layout): Update for
20651         location_get_source_line returning a char_span.
20652         (struct char_span): Move to input.h.
20653         (struct correction): Update for fields in char_span becoming
20654         private.
20655         (struct source_line): Update for location_get_source_line
20656         returning a char_span.
20657         (layout::print_line): Likewise.
20658         * edit-context.c (edited_file::print_content): Likewise.
20659         (edited_file::print_diff_hunk): Likewise.
20660         (edited_file::print_run_of_changed_lines): Likewise.
20661         (edited_file::get_num_lines): Likewise.
20662         (edited_line::edited_line): Likewise.
20663         * final.c (asm_show_source): Likewise.
20664         * input.c (location_get_source_line): Convert return type
20665         from const char * to char_span, losing the final "line_len"
20666         param.
20667         (dump_location_info): Update for the above.
20668         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20669         when handling the literal within the line.
20670         (test_reading_source_line): Update for location_get_source_line
20671         returning a char_span.
20672         * input.h (class char_span): Move here from
20673         diagnostic-show-locus.c, converting from a struct to a class.
20674         Make data members private.
20675         (char_span::operator bool): New.
20676         (char_span::length): New.
20677         (char_span::get_buffer): New.
20678         (char_span::operator[]): New.
20679         (char_span::subspan): Make const.
20680         (char_span::xstrdup): New.
20681         (location_get_source_line): Convert return type from const char *
20682         to char_span, losing the final "line_size" param.
20684 2018-04-30  Jan Hubicka  <jh@suse.cz>
20686         * lto-wrapper.c (ltrans_priorities): New static var.
20687         (cmp_priority): New.
20688         (run_gcc): Read priorities and if doing parallel build order
20689         the Makefile by them.
20691 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20693         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20695 2018-04-30  Richard Biener  <rguenther@suse.de>
20697         * tree-cfg.c (verify_address): Remove base argument, add
20698         flag whether to check TREE_ADDRESSABLE and do that.
20699         (verify_expr): Remove.
20700         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20701         (verify_gimple_assign_single): Likewise.
20702         (verify_gimple_switch): Likewise.
20703         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20704         piece from verify_expr.
20705         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20707 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20709         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20711 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20713         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20714         (small_data_pattern): Likewise.
20715         (arc_rewrite_small_data): Likewise.
20716         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20717         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20718         (get_symbol_alignment): New function.
20719         (legitimate_small_data_address_p): Likewise.
20720         (legitimate_scaled_address): Update, call
20721         legitimate_small_data_address_p.
20722         (output_sdata): New static variable.
20723         (arc_print_operand): Update how we handle small data operands.
20724         (arc_print_operand_address): Likewise.
20725         (arc_legitimate_address_p): Update, use
20726         legitimate_small_data_address_p.
20727         (arc_rewrite_small_data_p): Remove.
20728         (arc_rewrite_small_data_1): Likewise.
20729         (arc_rewrite_small_data): Likewise.
20730         (small_data_pattern): Likewise.
20731         (compact_sda_memory_operand): Update to use
20732         legitimate_small_data_address_p and get_symbol_alignment.
20733         (prepare_move_operands): Don't rewite sdata pattern.
20734         (prepare_extend_operands): Remove.
20735         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20736         pattern.
20737         (zero_extendqisi2): Likewise.
20738         (zero_extendhisi2): Likewise.
20739         (extendqihi2): Likewise.
20740         (extendqisi2): Likewise.
20741         (extendhisi2): Likewise.
20742         (addsi3): Likewise.
20743         (subsi3): Likewise.
20744         (andsi3): Likewise.
20745         * config/arc/constraints.md (Usd): Change it to memory constraint.
20747 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20749         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20750         as source of std instructions.
20751         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20752         allow 6-bit constants as source for store instructions.
20753         (movdi_insn): Update instruction pattern to allow 6-bit constants
20754         as source for store instructions.
20756 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20758         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20760 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20761             Sandra Loosemore <sandra@codesourcery.com>
20763         * dumpfile.c (dump_open): Allow '-' for stdout.
20764         * doc/invoke.texi (Developer Options): Document dump filename
20765         determination early.  Document stdin/stdout selection.
20767 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20769         Microblaze Target: PIC data text relative
20771         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20772         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20773         Add declaration.
20774         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20775         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20776         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20777         New addressing mode for data-text relative position indepenedent code.
20778         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20779         'ADDRESS_SYMBOLIC_TXT_REL'.
20780         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20781         (microblaze_legitimate_pic_operand): Exclude function calls from
20782         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20783         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20784         addresses cases.
20785         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20786         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20787         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20788         for 'address + offset'.
20789         (microblaze_expand_prologue): Add new function prologue call for
20790         'r20' assignation.
20791         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20792         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20793         table in case of TARGET_PIC_DATA_TEXT_REL.
20794         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20795         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20796         Add new macros 'UNSPEC_TEXT',
20797         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20798         + exclude function calls from 'UNSPEC_PLT' in case of data text
20799         relative mode.
20800         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20801         new target hook for generating address diff vector tables in case of
20802         flag_pic.
20803         * doc/tm.texi : Regenerate.
20804         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20805         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20806         of addr diff vector generation.
20807         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20808         target hook definition.
20809         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20810         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20811         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20812         option for data text relative.
20814 2018-04-30  Richard Biener  <rguenther@suse.de>
20816         * tree-chrec.h (evolution_function_is_constant_p): Remove
20817         redundant check.
20818         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20820 2018-04-30  Richard Biener  <rguenther@suse.de>
20822         PR bootstrap/85571
20823         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20825 2018-04-30  Richard Biener  <rguenther@suse.de>
20827         PR tree-optimization/28364
20828         PR tree-optimization/85275
20829         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20830         copying first exit test.
20832 2018-04-28  Mark Wielaard  <mark@klomp.org>
20834         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20835         dwarf_version >= 5.
20836         (dwarf_AT): Handle DW_AT_addr_base.
20837         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20839 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20841         PR target/84431
20842         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20843         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20844         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20845         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20847 2018-04-28  Richard Biener  <rguenther@suse.de>
20849         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20850         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20851         to reflect use.  Only add interesting stmts.
20853 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20855         PR ipa/85549
20856         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20857         the jump function allows for passing through aggregate values.
20859 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20861         * input.h (in_system_header_at): Convert from macro to inline
20862         function.
20863         (from_macro_expansion_at): Likewise.
20864         (from_macro_definition_at): Likewise.
20866 2018-04-27  Jeff Law  <law@redhat.com>
20868         * config.gcc: Mark tile* targets as deprecated/obsolete.
20870 2018-04-27  Richard Biener  <rguenther@suse.de>
20872         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20873         fix for ILP32.
20875 2018-04-27  Richard Biener  <rguenther@suse.de>
20877         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20879 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
20881         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20882         with Yd constraint. Set "preferred_for_speed" attribute from
20883         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20884         with Yd constraint.
20885         (*movdi_internal): Ditto.
20886         (movti_interunit splitters): Remove
20887         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20888         (movdi_interunit splitters): Ditto.
20889         * config/i386/constraints.md (Ye): Remove.
20890         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20892 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20894         PR target/85512
20895         * config/aarch64/constraints.md (Usg): Limit to 31.
20896         (Usj): Limit to 63.
20898 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
20900         PR tree-optimization/85529
20901         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20902         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20903         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20904         zero extension or masking of the MSB bit.
20905         (optimize_range_tests): Add FIRST_BB argument, pass it through
20906         to optimize_range_tests_var_bound.
20907         (maybe_optimize_range_tests, reassociate_bb): Adjust
20908         optimize_range_tests callers.
20910 2018-04-26  Richard Biener  <rguenther@suse.de>
20911             Jakub Jelinek  <jakub@redhat.com>
20913         * cgraph.h (symbol_table): Just declare debug method here.
20914         * symtab.c (symbol_table::debug): Define.
20916 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
20918         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20920 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
20922         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20923         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20924         (*movdi_internal): Substitute Yi and Yj constraint with x
20925         and Ym and Yn constraint with y constraint.  Update "isa"
20926         attribute and set "preferred_for_speed" attribute from
20927         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20928         (*movsi_internal): Ditto.
20929         (*movdf_internal): Ditto.
20930         (*movsf_internal): Ditto.
20931         (*zero_extendsidi2): Ditto.
20932         * config/i386/sse.md (vec_set<mode>_0): Ditto.
20933         (sse2_loadld): Ditto.
20934         (*vec_extract<ssevecmodelower>_0): Ditto.
20935         (*vec_extractv4si_0_zext_sse4): Ditto.
20936         (vec_concatv2di): Ditto.
20937         (*vec_dup<mode>): Ditto.
20938         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20939         * config/i386/constraints.md (Yi): Remove.
20940         (Yj): Remove.
20941         (Ym): Remove.
20942         (Yn): Remove.
20944 2018-04-26  Nathan Sidwell  <nathan@acm.org>
20946         * dumpfile.c (dump_open): New.
20947         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20948         (dump_finish): Detect stdio/stderr by value not name.
20950 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
20952         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20954 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20956         PR target/84952
20957         * config/nvptx/nvptx.c (verify_neutering_jumps)
20958         (verify_neutering_labels): New function
20959         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20961 2018-04-26  Tom de Vries  <tom@codesourcery.com>
20963         PR target/84025
20964         * config/nvptx/nvptx.c (needs_neutering_p): New function.
20965         (nvptx_single): Use needs_neutering_p to skip over insns that do not
20966         need neutering.
20968 2018-04-26  Richard Biener <rguenther@suse.de>
20969             Tom de Vries  <tom@codesourcery.com>
20971         PR lto/85422
20972         * lto-streamer-out.c (output_function): Fixup loops if required to match
20973         discovery done in the reader.
20975 2018-04-26  Richard Biener  <rguenther@suse.de>
20977         PR tree-optimization/85116
20978         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20979         have a loop exit from the single latch predecessor.  Remove
20980         case of header with just condition.
20981         (ch_base::copy_headers): Exclude infinite loops from any
20982         processing.
20983         (pass_ch::execute): Record exits.
20985 2018-04-26  Richard Biener  <rguenther@suse.de>
20987         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20988         prologue cost vector and pass it to vect_get_load_cost.
20989         (vect_get_peeling_costs_all_drs): Likewise.
20990         (vect_peeling_hash_get_lowest_cost): Likewise.
20991         (vect_enhance_data_refs_alignment): Likewise.
20993 2018-04-26  Richard Biener  <rguenther@suse.de>
20995         PR middle-end/85450
20996         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20997         checking of integer<->pointer conversions.
20998         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20999         sign-/zero-extending pointer types.
21000         (expand_omp_for_static_chunk): Likewise.
21002 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21003             Jean Lee  <xiaoyur347@gmail.com>
21005         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21006         (TARGET_ASAN_SHADOW_OFFSET): Define.
21007         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21008         true for -fsanitize=address.
21010 2018-04-25  Mark Wielaard  <mark@klomp.org>
21012         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21013         shorter ones.
21015 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21017         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21018         than "alu", remove explicit "memory" and "imm_disp" attributes.
21019         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21021         PR middle-end/85414
21022         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21023         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21024         gen_lowpart_no_emit.
21026 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21028         PR target/85473
21029         * config/i386/i386.c (ix86_expand_builtin): Change memory
21030         operand to XI, extend p0 to Pmode.
21031         * config/i386/i386.md: Change unspec volatile and operand
21032         1 mode to XI, change operand 0 mode to P.
21034 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21036         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21037         GET_MODE_MASK before any checking.
21038         (nds32_can_use_bset_p): Likewise.
21039         (nds32_can_use_btgl_p): Likewise.
21041 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21043         * config/nds32/nds32-doubleword.md: New define_split pattern for
21044         illegal register number.
21046 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21048         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21050 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21052         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21054 2018-04-25  Richard Biener  <rguenther@suse.de>
21056         * lto-streamer.h (LTO_major_version): Bump to 8.
21058 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21060         * BASE-VER: Set to 9.0.0.
21062 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21064         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21065         in __abskf2 and __powikf2.
21067 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21069         PR target/85512
21070         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21071         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21072         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21073         Use the above on operand 2.  Reindent.
21074         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21076 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21078         PR target/85485
21079         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21080         handle OPT_mcet.
21081         * config/i386/i386.opt (mcet): Removed.
21082         * doc/install.texi: Remove -mcet documentation.
21083         * doc/invoke.texi: Likewise.
21085 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21087         PR target/85485
21088         * doc/install.texi: Remove -mcet from bootstrap-cet.
21090 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21092         PR target/85511
21093         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21094         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21095         if TARGET_64BIT.
21097         PR target/85503
21098         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21099         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21100         containing a CONST_VECTOR.
21102 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21104         * doc/install.texi: Update newlib dependency for nvptx.
21106 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21108         PR target/85508
21109         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21110         instead of INTVAL when shifting x left.
21112 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21114         PR tree-optimization/85478
21115         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21116         vect_grouped_store_supported for single element vectors.
21118 2018-04-24  Richard Biener  <rguenther@suse.de>
21120         PR target/85491
21121         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21122         load cost increase to the case of non-constant step.
21124 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21126         PR target/84828
21127         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21128         destination if any_malformed_asm.
21130 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21132         PR middle-end/85496
21133         * expr.c (store_field): In the bitfield case, if the value comes from
21134         a function call and is returned in registers by means of a PARALLEL,
21135         do not change the mode of the temporary unless BLKmode and VOIDmode.
21137 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21139         PR rtl-optimization/85423
21140         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21141         dependencies to debug insns when the previous insn is non-debug.
21143 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21145         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21146         enums into a single definition.
21147         (fls): Fix predicates and printing.
21148         (seti): Likewise.
21150 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21152         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21153         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21154         and short u6 immediate.
21155         (check_if_valid_sleep_operand): Remove.
21156         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21158 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21160         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21161         flag_always_save_lp condition.
21162         * config/nds32/nds32.opt (malways-save-lp): New option.
21164 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21166         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21167         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21168         * config/nds32/nds32.h
21169         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21170         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21172 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21174         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21175         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21177 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21178             Chung-Ju Wu  <jasonwucj@gmail.com>
21180         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21181         Declare.
21182         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21183         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21185 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21187         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21189 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21191         * config/nds32/nds32-protos.h (nds32_data_alignment,
21192         nds32_local_alignment): Declare.
21193         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21194         nds32_local_alignment): New functions.
21195         (TARGET_CONSTANT_ALIGNMENT): Define.
21196         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21198 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21200         * config/nds32/nds32.c
21201         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21202         (TARGET_MODES_TIEABLE_P): Likewise.
21204 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21206         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21207         level Ofast and Og.
21209 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21210             Chung-Ju Wu  <jasonwucj@gmail.com>
21212         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21213         for unaligned access.
21214         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21215         unaligned access.
21216         * config/nds32/nds32-intrinsic.md: Likewise.
21217         * config/nds32/nds32_intrinsic.h: Likewise.
21218         * config/nds32/nds32.h (nds32_builtins): Likewise.
21219         * config/nds32/nds32.opt (munaligned-access): New option.
21220         * config/nds32/nds32.c (nds32_asm_file_start): Display
21221         flag_unaligned_access status.
21223 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21225         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21226         -mno-relax is present.
21227         * config/riscv/linux.h (LINK_SPEC): Ditto.
21229 2018-04-20  Martin Sebor  <msebor@redhat.com>
21231         PR c/85365
21232         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21233         for null pointers.
21234         (gimple_fold_builtin_stxcpy_chk): Same.
21235         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21237 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21239         PR target/85456
21240         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21241         __powikf2 when long double is IEEE 128-bit.
21243 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21245         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21246         step to make sure stack always aligned.
21248 2018-04-20  Carl Love  <cel@us.ibm.com>
21250         PR target/83402
21251         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21252         size check for arg0.
21254 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21255             Tom de Vries  <tom@codesourcery.com>
21257         PR target/85445
21258         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21259         Emit insns for calls too.
21260         (nvptx_find_par): Always look for worker-level predecessor insn.
21261         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21262         calls.
21263         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21264         (nvptx_process_pars): Propagate frames for calls.
21266 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21268         PR target/85469
21269         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21270         Removed.
21271         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21272         (ix86_handle_option): Don't handle OPT_mibt.
21273         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21274         __SHSTK__.
21275         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21276         has_ibt and ibt.
21277         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21278         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21279         (ix86_target_macros): Define __CET__ with flag_cf_protection
21280         for -fcf-protection.
21281         * config/i386/i386.c (isa2_opts): Remove -mibt.
21282         * config/i386/i386.h (TARGET_IBT): Removed.
21283         (TARGET_IBT_P): Likewise.
21284         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21285         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21286         * config/i386/i386.opt (mcet): Update help message.
21287         (mshstk): Likewise.
21288         (mibt): Removed.
21289         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21290         -mcet as an alias for -mshstk.
21292 2018-04-20  Richard Biener <rguenther@suse.de>
21294         PR middle-end/85475
21295         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21296         complexity by forcing a single use of the multiply operand.
21298 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21300         ipa/85449
21301         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21302         recursion dependency to only apply to non-clones.
21304 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21306         ipa/85447
21307         * ipa-cp.c (create_specialized_node): Check that clones of
21308         self-recursive edges exist during IPA-CP.
21310 2018-04-19  Toon Moene  <toon@moene.org>
21312         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21313         by -O3.
21315 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21317         PR tree-optimization/85467
21318         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21319         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21320         VECTOR_CST element to type.
21322 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21324         PR target/85397
21325         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21326         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21327         (builtin_longjmp): Likewise.
21328         (save_stack_nonlocal): New pattern.
21329         (restore_stack_nonlocal): Likewise.
21331 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21333         PR target/85404
21334         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21335         Replace ASM_OUTPUT_LABEL with fprintf.
21337 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21339         PR target/85417
21340         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21341         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21342         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21343         define __IBT__ and __SHSTK__ for -fcf-protection.
21344         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21345         TARGET_IBT.
21346         (ix86_trampoline_init): Likewise.
21347         (x86_output_mi_thunk): Likewise.
21348         (ix86_notrack_prefixed_insn_p): Likewise.
21349         (ix86_option_override_internal): Don't disallow -fcf-protection.
21350         * config/i386/i386.md (rdssp<mode>): Also enable for
21351         -fcf-protection.
21352         (incssp<mode>): Likewise.
21353         (nop_endbr): Likewise.
21354         * config/i386/i386.opt (mcet): Change help message to built-in
21355         functions only.
21356         (mibt): Likewise.
21357         (mshstk): Likewise.
21358         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21359         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21360         enable CET built-in functions.
21362 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21364         * common/config/i386/i386-common.c
21365         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21366         OPTION_MASK_ISA_MOVDIRI_UNSET,
21367         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21368         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21369         * config.gcc (movdirintrin.h): New header.
21370         * config/i386/cpuid.h (bit_MOVDIRI,
21371         bit_MOVDIR64B): New bits.
21372         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21373         and -mmvodir64b.
21374         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21375         (VOID, PVOID, PCVOID)): New function types.
21376         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21377         __builtin_ia32_directstoreu_u64,
21378         __builtin_ia32_movdir64b): New builtins.
21379         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21380         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21381         and -mmovdiri.
21382         (ix86_valid_target_attribute_inner_p): Ditto.
21383         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21384         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21385         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21386         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21387         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21388         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21389         (movdiri<mode>, movdir64b_<mode>): New.
21390         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21391         * config/i386/immintrin.h: Include movdirintrin.h.
21392         * config/i386/movdirintrin.h: New file.
21393         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21395 2018-04-19  Richard Biener  <rguenther@suse.de>
21397         PR middle-end/85455
21398         * cfg.c (clear_bb_flags): When loop state says we have
21399         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21401 2018-04-19  Richard Biener  <rguenther@suse.de>
21403         PR tree-optimization/84737
21404         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21405         copying restrict info.
21406         (vect_setup_realignment): Use it.
21407         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21408         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21409         the first DR to all generated stores.
21410         (vectorizable_load): Likewise for loads.
21412 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21414         PR tree-optimization/85446
21415         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21416         the integral and pointer types to have the same precision.
21418         * doc/install.texi: Document --disable-cet being the default and
21419         --enable-cet=auto.
21421 2018-04-18  Martin Liska  <mliska@suse.cz>
21423         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21424         style.
21426 2018-04-18  Martin Liska  <mliska@suse.cz>
21428         Revert
21429         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21431         PR ipa/83983
21432         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21433         arguments if they are comparable.
21435 2018-04-18  Martin Liska  <mliska@suse.cz>
21437         Revert
21438         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21440         PR lto/84805
21441         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21442         incomplete types.
21444 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21446         PR target/85388
21447         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21448         ENDBR after calling __morestack.
21450 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21452         PR jit/85384
21453         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21454         by using gcc_base_ver to generate a gcc_driver_version, and use
21455         it when generating GCC_DRIVER_NAME.
21456         * configure: Regenerate.
21458 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21460         PR target/81084
21461         * config.gcc: Obsolete powerpc*-*-*spe*.
21463 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21465         PR debug/84637
21466         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21467         (stabstr_D): Change type of unum from unsigned int to
21468         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21469         type.
21471 2018-04-17  Jim Wilson  <jimw@sifive.com>
21473         PR 84856
21474         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21475         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21476         Set arg_pointer_offset after using pretend_args_size.
21478 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21480         PR rtl-optimization/85431
21481         * dse.c (record_store): Ignore zero width stores.
21483         PR sanitizer/85230
21484         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21485         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21486         __builtin_stack_restore rather than after it.
21487         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21488         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21489         argument instead of virtual_dynamic_stack_rtx.
21491 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21493         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21494         New prototype.
21495         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21496         Add note to error message to explain internal mapping of overloaded
21497         built-in function name to non-overloaded built-in function name.
21498         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21499         function.
21501 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21503         PR target/85424
21504         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21505         where the inputs overlap with the output.
21507 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21509         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21510         (=v, v) alternative and explicit "memory" attribute.
21511         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21512         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21513         attributes.
21514         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21515         "sselog1" type instead of "sselog".
21516         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21517         "sselog".  Remove explicit "memory" attribute.
21518         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21519         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21520         attributes.
21521         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21522         "sselog1" type instead of "sselog".  Remove explicit "memory"
21523         attribute.
21524         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21525         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21526         attribute.
21527         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21528         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21529         attributes.
21530         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21531         "sselog1" type instead of "sselog".  Remove explicit "memory"
21532         attribute.
21533         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21534         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21535         attribute.
21537         PR target/85430
21538         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21540         PR middle-end/85414
21541         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21542         on a SUBREG.
21544 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21546         PR ipa/85421
21547         * ipa-cp.c (create_specialized_node): Call
21548         expand_all_artificial_thunks if necessary.
21550 2018-04-17  Martin Liska  <mliska@suse.cz>
21552         PR lto/85405
21553         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21554         in message, remote space in between '_G' and '('.
21556 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21558         PR target/85281
21559         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21560         avx512f_vmcmp<mode>3<round_saeonly_name>,
21561         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21562         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21563         avx512f_rndscale<mode><round_saeonly_name>,
21564         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21565         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21566         Use %<iptr>2 instead of %2 for -masm=intel.
21567         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21568         avx512f_vcvttss2usi<round_saeonly_name>,
21569         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21570         -masm=intel.
21571         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21572         avx512f_vcvttsd2usi<round_saeonly_name>,
21573         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21574         Use %q1 instead of %1 for -masm=intel.
21575         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21576         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21577         of %3 for -masm=intel.
21578         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21579         -masm=intel.
21580         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21581         -masm=intel.
21582         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21583         -masm=intel.
21584         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21585         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21586         %g1.
21587         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21588         -masm=intel.
21589         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21590         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21591         %g1 and one with %0 and %1.
21592         (avx512er_vmrcp28<mode><round_saeonly_name>,
21593         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21594         %1 for -masm=intel.
21595         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21596         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21597         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21598         of %0 and %{%4%} for -masm=intel.
21599         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21600         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21601         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21602         order of %0 and %{%5%}%{z%} for -masm=intel.
21604 2018-04-17  Jan Hubicka  <jh@suse.cz>
21606         PR lto/85405
21607         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21609 2018-04-17  Martin Liska  <mliska@suse.cz>
21611         PR ipa/85329
21612         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21613         for target_clone error message.  Make default implementation
21614         clone to be a local declaration.
21615         (separate_attrs): Add new argument and check for an empty
21616         string.
21617         (expand_target_clones): Handle it.
21618         (ipa_target_clone): Make redirection just for target_clones
21619         functions.
21621 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21622             Tom de Vries  <tom@codesourcery.com>
21624         PR middle-end/84955
21625         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21626         tiled basic blocks without omp continue statements.
21628 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21630         PR target/83660
21631         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21632         vec_extract expression as having side effects to make sure it gets
21633         a cleanup point.
21635 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21637         PR target/85403
21638         * config/i386/i386.c (get_builtin_code_for_version): Check
21639         error_mark_node.
21641 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21643         PR target/84331
21644         * config.gcc: Support "skylake".
21645         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21646         PROCESSOR_SKYLAKE.
21647         * config/i386/i386.c (m_SKYLAKE): Define.
21648         (processor_target_table): Add "skylake".
21649         (ix86_option_override_internal): Add "skylake".
21650         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21651         PROCESSOR_CANNONLAKE.
21652         (get_builtin_code_for_version): Fix priority for
21653         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21654         PROCESSOR_SKYLAKE-AVX512.
21655         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21656         (processor_type): Add PROCESSOR_SKYLAKE.
21658 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21659             Jason Merrill  <jason@redhat.com>
21661         PR c++/85112
21662         * convert.c (convert_to_integer_1): Use direct recursion for
21663         enumeral types and types with a precision less than the number
21664         of bits in their mode.
21666 2018-04-16  Julia Koval  <julia.koval@intel.com>
21668         PR target/84413
21669         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21670         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21672 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21674         PR target/85293
21675         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21676         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21677         and -mno-direct-move.
21679 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21681         PR target/83402
21682         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21683         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21684         Ensure negative shifts result in {0}.
21686 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21688         PR rtl-optimization/79916
21689         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21690         regs (if any) to define how to gnerate SD moves when LRA is in
21691         progress.
21693 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21695         PR rtl-optimization/85393
21696         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21697         * except.c (expand_dw2_landing_pad_for_region): Make static.
21698         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21699         a label and unconditional jump to old_bb, rather than
21700         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21701         basic block.
21703         PR rtl-optimization/85376
21704         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21705         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21706         instead of a specific value.
21708 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21709             Bin Cheng  <bin.cheng@arm.com>
21711         PR tree-optimization/82965
21712         PR tree-optimization/83991
21713         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21714         by_profile_only parameter.
21715         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21716         information if the loop was predicted to iterate too many times.
21717         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21719 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21721         PR lto/71991
21722         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21723         always inline.
21725 2018-04-13  Martin Liska  <mliska@suse.cz>
21726             Jakub Jelinek  <jakub@redhat.com>
21728         PR middle-end/81657
21729         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21730         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21731         * builtins.c (expand_builtin_memory_copy_args): Use
21732         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21733         handle dest_addr == pc_rtx.
21735 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21737         PR target/85291
21738         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21739         asked to not generate direct moves.
21740         (fix_trunc<mode>si2_stfiwx): Similar.
21741         (fix_trunc<mode>si2_internal): Similar.
21743 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21745         PR debug/83157
21746         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21747         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21748         lookup if dest in some wider mode is known to be const0_rtx and
21749         if so, record permanent equivalence for it to be ZERO_EXTEND of
21750         the narrower mode destination.
21752 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21754         * lto-streamer-out.c (output_function): Revert 259346.
21755         * omp-expand.c (expand_oacc_for): Likewise.
21757 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21759         PR rtl-optimization/85354
21760         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21761         * sel-sched.c (sel_global_init): ... here.
21763 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21765         PR target/85238
21766         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21767         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21768         mode for PE-COFF targets.
21769         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21770         (i386_pe_asm_lto_end): Likewise.
21771         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21772         (TARGET_ASM_LTO_END): Likewise.
21773         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21774         (i386_pe_asm_lto_start): New function.
21775         (i386_pe_asm_lto_end): Likewise.
21777 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21778             Richard Biener  <rguenther@suse.de>
21780         PR middle-end/84955
21781         * lto-streamer-out.c (output_function): Fix CFG loop state before
21782         streaming out.
21783         * omp-expand.c (expand_oacc_for): Handle calls to internal
21784         functions like regular functions.
21786 2018-04-12  Richard Biener  <rguenther@suse.de>
21788         PR lto/85371
21789         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21790         for the early LTO debug to properly generate references to it
21791         during DIE emission.  Do not re-use that for the skeleton for
21792         split-dwarf.
21793         (dwarf2out_early_finish): Likewise.
21795 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21797         PR target/85328
21798         * config/i386/sse.md
21799         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21800         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21801         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21802         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21803         and output is a reg, avoid creating invalid lowpart subreg, but
21804         instead split into a 512-bit move.  Don't split if not AVX512VL,
21805         input is xmm16+ reg and output is a mem.
21806         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21807         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21808         xmm16+ reg and output is a mem.
21810 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21812         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21813         also for flag_dwarf2_cfi_asm.
21815 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21817         PR rtl-optimization/85342
21818         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21819         a bool scalar var inside of the loop instead.  Don't try to update
21820         recog_data.operand after failed apply_change_group.
21822 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21824         PR target/85296
21825         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21826         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21827         array with flexible array member as array without given dimension.
21828         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21829         argument for undefined param to true.
21831 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21833         PR target/85321
21834         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21835         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21836         from PowerPC section.
21837         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21838         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21839         help text that is too long.
21840         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21841         help text that is too long.
21842         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21843         help text that is too long.
21845 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21847         * config/alpha/alpha.md (stack_probe_internal): Rename
21848         from "probe_stack".  Update all callers.
21850 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21852         PR rtl-optimization/84566
21853         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21854         sched_macro_fuse_insns.
21856 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21858         PR target/84301
21859         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21860         (compute_block_dependences): ... from here.
21862 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21864         PR tree-optimization/85331
21865         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21866         from int to HOST_WIDE_INT.
21868 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21870         PR ipa/84149
21871         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21872         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21873         not the same as the source val.
21874         (cgraph_edge_brings_value_p): New parameter.
21875         (gather_edges_for_value): Pass destination value to
21876         cgraph_edge_brings_value_p.
21877         (perhaps_add_new_callers): Likewise.
21878         (get_info_about_necessary_edges): Likewise and exclude values brought
21879         only by self-recursive edges.
21880         (create_specialized_node): Redirect only clones of self-calling edges.
21881         (+self_recursive_pass_through_p): New function.
21882         (find_more_scalar_values_for_callers_subset): Use it.
21883         (find_aggregate_values_for_callers_subset): Likewise.
21884         (known_aggs_to_agg_replacement_list): Removed.
21885         (decide_whether_version_node): Re-calculate known constants for all
21886         remaining context clones.
21888 2018-04-11  Richard Biener  <rguenther@suse.de>
21890         PR lto/85339
21891         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21892         from early DWARF output.
21893         (dwarf2out_early_finish): Output line info unconditionally into
21894         early DWARF and add reference to it.
21896 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21898         PR target/85281
21899         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21900         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21901         other than V2DFmode using iptr mode attribute.
21902         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21904 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21906         PR rtl-optimization/84659
21907         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21909 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21911         PR debug/85302
21912         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21913         SIZEP is NULL.
21914         (output_loc_list): Pass address of a dummy size variable even in the
21915         locview handling loop.
21916         (index_location_lists): Add comment on why skip_loc_list_entry can't
21917         call size_of_locs.
21919 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21921         PR target/85261
21922         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21923         into register.
21925 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
21927         PR target/85321
21928         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21929         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21930         and -mstring-compare-inline-limit.
21932 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21934         PR target/85287
21935         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21936         for stack clash protection in a register whenever we need it to be in
21937         a register.
21939 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21941         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21942         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21944 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
21946         PR target/85321
21947         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21948         the help text.
21949         (mlong-double-): Ditto.
21950         * config/rs6000/sysv4.opt (msdata=): Ditto.
21951         (mtls-size=): Ditto.
21953 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21955         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21956         erroneous entries for
21957         "vector int vec_ldl (int, long int *)", and
21958         "vector unsigned int vec_ldl (int, unsigned long int *)".
21959         Add comments and entries for
21960         "vector bool char vec_ldl (int, bool char *)",
21961         "vector bool short vec_ldl (int, bool short *)",
21962         "vector bool int vec_ldl (int, bool int *)",
21963         "vector bool long long vec_ldl (int, bool long long *)",
21964         "vector pixel vec_ldl (int, pixel *)",
21965         "vector long long vec_ldl (int, long long *)",
21966         "vector unsigned long long vec_ldl (int, unsigned long long *)".
21967         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21968         type tree bool_long_long_type_node and correct definition of
21969         bool_V2DI_type_node to make reference to this new type tree.
21970         (rs6000_mangle_type): Replace erroneous reference to
21971         bool_long_type_node with bool_long_long_type_node.
21972         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21973         comments to emphasize sign distinctions for char and int types and
21974         replace RS6000_BTI_bool_long constant with
21975         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
21976         use of RS6000_BTI_pixel.
21977         (bool_long_type_node): Remove this macro definition.
21978         (bool_long_long_type_node): New macro definition
21980 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21982         PR rtl-optimization/85300
21983         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21984         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21985         simplify_unary_operation fails.
21987 2018-04-10  Martin Liska  <mliska@suse.cz>
21989         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21990         cgraph_edge and ipa_ref.
21992 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
21994         PR target/85177
21995         PR target/85255
21996         * config/i386/sse.md
21997         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21998         computation of the VEC_MERGE selector from mask.
21999         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22000         Fix decoding of the VEC_MERGE selector into mask.
22002 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22004         PR tree-optimization/85286
22005         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22007 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22009         * final.c (final_1): Set insn_last_address as well as
22010         insn_current_address.
22012 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22014         PR target/85173
22015         * explow.c (emit_stack_probe): Call validize_mem on memory location
22016         before passing it to gen_probe_stack.  Create address operand and
22017         legitimize it for the probe_stack_address case.
22019 2018-04-09  Jan Hubicka  <jh@suse.cz>
22021         PR lto/85078
22022         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22023         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22024         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22025         walk basetypes.
22026         (free_lang_data): Rebuild type inheritance graph.
22028 2018-04-09  Martin Sebor  <msebor@redhat.com>
22030         * invoke.texi (-finline-small-functions): Mention other optimization
22031         options.
22032         (-findirect-inlining, -fpartial-inlining): Same.
22033         (-finline-functions-called-once): Same.
22034         (-freorder-blocks-and-partition): Same.
22036 2018-04-09  Jan Hubicka  <jh@suse.cz>
22038         PR rtl/84058
22039         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22040         jumps; choose last target that matches the criteria (i.e.
22041         no partition changes for non-crossing jumps).
22042         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22043         support for redirecting crossing jumps to non-crossing.
22045 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22047         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22048         also for naked functions.
22050 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22052         * config/arc/arc.md (add_shift): New pattern.
22053         (add_shift2): Likewise.
22054         (sub_shift): Likewise.
22055         (sub_shift_cmp0_noout): Likewise.
22056         (compare_si_ashiftsi): Likewise.
22057         (xbfu_cmp0_noout): New combine pattern.
22058         (xbfu_cmp0"): Likewise.
22059         (movsi_set_cc_insn): Place the predicable variant first.
22060         (commutative_binary_cmp0_noout): Remove clobber.
22061         (commutative_binary_cmp0): New pattern.
22062         (noncommutative_binary_cmp0): Likewise.
22063         (noncommutative_binary_cmp0_noout): Likewise.
22064         (noncommutative_binary_comparison_result_used): Removed.
22065         (rsub_cmp0): New pattern.
22066         (rsub_cmp0_noout): Likewise.
22067         (extzvsi): Changed, keep only meaningful variants.
22068         (SQH, SEZ): New iterators.
22069         (SQH_postfix): New mode attribute.
22070         (SEZ_prefix): New code attribute.
22071         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22072         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22073         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22074         of numerical value.
22075         (noncommutative_operator): Check the availability of barrel
22076         shifter option.
22078 2018-04-09  Richard Biener  <rguenther@suse.de>
22080         PR tree-optimization/85284
22081         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22082         Only use the niter constraining form of simple_iv when the exit
22083         is always executed.
22085 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22087         PR target/84041
22088         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22089         (define_expand "*memory_barrier"): New define_expand.
22090         (define_insn "memory_barrier"): New insn.
22092 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22094         PR rtl-optimization/80463
22095         PR rtl-optimization/83972
22096         PR rtl-optimization/83480
22098         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22099         correct producer for the insn.
22100         (tidy_control_flow): Fixup seqnos in case of debug insns.
22102 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22104         PR rtl-optimization/83913
22106         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22107         different sched-times when merging exprs.
22109 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22111         PR rtl-optimization/83962
22113         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22114         tidy_fallthru_edge and tidy_control_flow.
22116 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22118         PR rtl-optimization/83530
22120         * sel-sched.c (force_next_insn): New global variable.
22121         (remove_insn_for_debug): When force_next_insn is true, also leave only
22122         next insn in the ready list.
22123         (sel_sched_region): When the region wasn't scheduled, make another pass
22124         over it with force_next_insn set to 1.
22126 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22128         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22129         into tm_file.
22130         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22131         for interrupt control.
22132         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22133         functions for interrupt control.
22134         * config/nds32/nds32-intrinsic.md: Likewise.
22135         * config/nds32/nds32_intrinsic.h: Likewise.
22136         * config/nds32/nds32.h (nds32_builtins): Likewise.
22138 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22140         * config/nds32/nds32.c (nds32_init_machine_status,
22141         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22142         strict_aligned_p field.
22143         (nds32_expand_to_rtl_hook): New function.
22144         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22145         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22147 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22148             Chung-Ju Wu  <jasonwucj@gmail.com>
22150         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22151         * config/nds32/nds32-n7.md: New file.
22152         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22153         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22154         pipeline.
22155         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22156         * config/nds32/nds32.md (pipeline_model): Add n7.
22157         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22158         * config/nds32/pipelines.md: Include n7 settings.
22160 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22161             Chung-Ju Wu  <jasonwucj@gmail.com>
22163         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22164         * config/nds32/nds32-e8.md: New file.
22165         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22166         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22167         pipeline.
22168         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22169         * config/nds32/nds32.md (pipeline_model): Add e8.
22170         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22171         * config/nds32/pipelines.md: Include e8 settings.
22173 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22174             Chung-Ju Wu  <jasonwucj@gmail.com>
22176         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22177         * config/nds32/nds32-n8.md: New file.
22178         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22179         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22180         pipeline.
22181         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22182         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22183         * config/nds32/nds32.md (pipeline_model): Add n8.
22184         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22185         * config/nds32/pipelines.md: Include n8 settings.
22187 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22188             Chung-Ju Wu  <jasonwucj@gmail.com>
22190         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22191         * config/nds32/nds32-n9-2r1w.md: New file.
22192         * config/nds32/nds32-n9-3r2w.md: New file.
22193         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22194         nds32_register_ports): New or modify for cpu n9.
22195         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22196         pipeline.
22197         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22198         * config/nds32/nds32-utils.c: New file.
22199         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22200         TARGET_MUL_SLOW): Define.
22201         * config/nds32/nds32.md (pipeline_model): New attribute.
22202         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22203         New options that support cpu n9.
22204         * config/nds32/pipelines.md: Include n9 settings.
22205         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22207 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22209         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22210         information if necessary.
22211         (output_cond_branch_compare_zero): Likewise.
22212         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22213         (nds32_target_alignment): Refine for alignment.
22214         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22215         (FUNCTION_BOUNDARY): Modify.
22216         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22217         align case.
22218         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22220 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22222         * config/nds32/constants.md (unspec_volatile_element): Add values for
22223         TLB operation and data prefetch.
22224         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22225         functions for TLB operation and data prefetch.
22226         * config/nds32/nds32-intrinsic.md: Likewise.
22227         * config/nds32/nds32_intrinsic.h: Likewise.
22228         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22229         (nds32_print_operand): Likewise.
22230         * config/nds32/nds32.h (nds32_builtins): Likewise.
22232 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22233         Andrew Pinski <pinsika@gcc.gnu.org>
22235         PR middle-end/82976
22236         * match.pd: Use constant_boolean_node of correct type instead of
22237         boolean_true_node or boolean_false_node for simplifying
22238         pointer comparisons to zero.
22240 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22242         PR tree-optimization/80021
22243         * tree.c (verify_type_variant): Make error call in verify_variant_match
22244         translatable and remove final full stop.
22246 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22248         * config/nds32/constants.md (unspec_volatile_element): Add
22249         UNSPEC_VOLATILE_EH_RETURN.
22250         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22251         nds32_output_stack_pop): Support dwarf exception handling process.
22252         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22253         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22254         exception handling process.
22255         (nds32_compute_stack_frame): Likewise.
22256         (nds32_return_addr_rtx): Likewise.
22257         (nds32_initial_elimination_offset): Likewise.
22258         (nds32_expand_prologue): Likewise.
22259         (nds32_expand_epilogue): Likewise.
22260         (nds32_dynamic_chain_address): New function.
22261         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22262         exception handling.
22263         (DYNAMIC_CHAIN_ADDRESS): Define.
22264         (EH_RETURN_DATA_REGNO): Define.
22265         (EH_RETURN_STACKADJ_RTX): Define.
22266         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22267         patterns for dwarf exception handling.
22269 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22271         * config/nds32/nds32.h: Clean up obsolete macros.
22273 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22275         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22276         Add enum values for particular instructions.
22277         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22278         particular intrinsic functions.
22279         * config/nds32/nds32-intrinsic.md: Likewise.
22280         * config/nds32/nds32_intrinsic.h: Likewise.
22281         * config/nds32/nds32.h (nds32_builtins): Likewise.
22282         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22283         (btst, ave): New patterns for particular instructions.
22285 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22287         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22288         Add enum values for atomic load/store and memory sync.
22289         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22290         and memory sync.
22291         * config/nds32/nds32-intrinsic.md: Likewise.
22292         * config/nds32/nds32_intrinsic.h: Likewise.
22293         * config/nds32/nds32.h (nds32_builtins): Likewise.
22295 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22297         PR tree-optimization/85257
22298         * fold-const.c (native_encode_vector): If not all elts could fit
22299         and off is -1, return 0 rather than offset.
22300         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22301         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22302         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22303         adjust buffer in native_interpret_expr call.
22305 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22307         * config/nds32/constants.md (unspec_volatile_element): Add cache
22308         control enum values.
22309         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22310         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22311         * config/nds32/nds32.c (nds32_cctl_names): New.
22312         (nds32_print_operand): Handle cache control register names.
22313         * config/nds32/nds32.h (nds32_builtins): New enum values.
22314         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22315         macros.
22316         * config/nds32/nds32.md (type): Add mmu.
22317         * config/nds32/pipelines.md (simple_insn): Add mmu.
22319 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22321         * config/nds32/nds32.md (type): Remove call.
22322         * config/nds32/pipelines.md (simple_insn): Likewise.
22324 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22326         * config/nds32/constants.md (unspec_volatile_element): Add
22327         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22328         UNSPEC_VOLATILE_FMFCFG.
22329         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22330         description for fmfcfg and fmfcsr.
22331         (bdesc_1arg): Add fmtcsr.
22332         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22333         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22334         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22335         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22336         unspec_fmfcfg): New patterns.
22337         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22338         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22339         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22340         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22341         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22342         __nds32__fmfcfg): Define.
22344 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22346         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22347         intrinsic register names.
22348         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22349         intrinsic register enum values and macros.
22351 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22353         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22354         for load/store addressing form.
22355         (nds32_print_operand_address): Likewise.
22357 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22359         PR target/85196
22360         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22361         based on LABEL_REF.  Remove useless assertion.
22362         (pic_address_needs_scratch): Fix formatting.
22363         (sparc_legitimize_pic_address): Minor tweaks.
22364         (sparc_delegitimize_address): Adjust assertion accordingly.
22365         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22366         into symbolic_operand.
22367         (movsi_high_pic_label_ref): Likewise.
22368         (movsi_lo_sum_pic_label_ref): Likewise.
22369         (movdi_pic_label_ref): Likewise.
22370         (movdi_high_pic_label_ref): Likewise.
22371         (movdi_lo_sum_pic_label_ref): Likewise.
22373 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22375         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22376         custom LIB_SPEC setup.
22378 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22379             Kito Cheng  <kito.cheng@gmail.com>
22381         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22382         * config/riscv/freebsd.h: New.
22384 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22386         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22387         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22388         file.
22390 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22391             Kito Cheng  <kito.cheng@gmail.com>
22393         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22394         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22395         * config/nds32/nds32-protos.h (nds32_output_call,
22396         nds32_output_return): Declare.
22397         * config/nds32/nds32.md: Refine all the call and return patterns.
22399 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22401         PR debug/85252
22402         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22403         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22405         PR rtl-optimization/84872
22406         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22407         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22408         EDGE_CROSSING edge.
22410 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22412         * expr.c (copy_blkmode_to_reg): Revert 254862.
22413         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22415 2018-04-06  Richard Biener  <rguenther@suse.de>
22417         PR middle-end/85244
22418         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22419         after seeing a component reference with an adjacent field.  Treat
22420         refs to arrays at struct end of external decls similar to
22421         refs to unconstrained commons.
22423 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22425         PR sanitizer/85213
22426         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22427         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22428         recursive calls.
22429         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22430         save_p here.
22432 2018-04-06  Richard Biener  <rguenther@suse.de>
22434         PR middle-end/85180
22435         * alias.c (find_base_term): New wrapper around find_base_term
22436         unwinding CSELIB_VAL_PTR changes.
22437         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22438         recursion.
22440 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22442         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22443         instructions.
22444         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22445         constant definitions.
22446         ("nop"): lr 0,0 -> nopr r0
22447         ("nop_lr0", "nop_lr1"): New insn definitions.
22449 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22451         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22452         NDS32_V3PUSH_AVAILABLE_P macro.
22454 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22455             Chung-Ju Wu  <jasonwucj@gmail.com>
22457         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22458         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22459         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22460         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22461         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22462         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22463         * config/nds32/constraints.md: New constraints and checking for hard
22464         float configuration.
22465         * config/nds32/iterators.md: New mode iterator and attribute for hard
22466         float configuration.
22467         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22468         patterns.
22469         * config/nds32/nds32-fpu.md: New file.
22470         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22471         deal with hard float code generation.
22472         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22473         ARCH_V3S.
22474         (abi_type, float_reg_number): New enum type.
22475         * config/nds32/nds32-predicates.c: New predicates for hard float.
22476         * config/nds32/nds32-protos.h: Declare functions for hard float.
22477         * config/nds32/nds32.c: Implementation for hard float configuration.
22478         * config/nds32/nds32.h: Definitions for hard float configuration.
22479         * config/nds32/nds32.md: Include hard float machine description and
22480         modify patterns for hard float configuration.
22481         * config/nds32/nds32.opt: New options for hard float configuration.
22482         * config/nds32/predicates.md: New predicates for hard float
22483         configuration.
22485 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22487         * common/config/nds32/nds32-common.c
22488         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22490 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22492         PR middle-end/85195
22493         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22494         CONSTRUCTOR_ELT (ctor, ...)->value.
22496 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22498         PR target/85193
22499         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22501 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22503         PR target/85204
22504         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22505         cond jump.
22507 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22508             Kito Cheng  <kito.cheng@gmail.com>
22510         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22511         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22512         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22513         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22515 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22516             Kito Cheng  <kito.cheng@gmail.com>
22518         * config/nds32/constraints.md (Ufe): New memory constraint.
22519         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22520         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22521         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22522         operands.
22523         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22524         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22526 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22528         * config/nds32/nds32.md: Use optimize_size in the condition for
22529         alu-shift instructions.
22531 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22533         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22535 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22537         * config/nds32/nds32.md (negsi2): Refine pattern.
22539 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22540             Chung-Ju Wu  <jasonwucj@gmail.com>
22542         * config/nds32/iterators.md (shift_rotate): New code iterator.
22543         (shift): New code attribute.
22544         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22545         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22546         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22547         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22548         bit-wise operations.
22549         (andsi3, *andsi3): Ditto.
22550         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22551         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22552         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22553         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22554         nds32_ior_operand, nds32_xor_operand): New predicates.
22556 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22558         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22559         (addsi3, subsi3): ... this.
22561 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22563         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22565 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22567         * config/nds32/nds32.md: Adjust indention.
22569 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22571         * config/nds32/nds32.md (feature): New attribute.
22573 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22575         * config/nds32/nds32.md (subtype): New attribute.
22577 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22579         PR target/85203
22580         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22581         expansion to perform a bitwise AND of the argument followed by a
22582         boolean negation of the result.
22584 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22586         PR rtl-optimization/84878
22587         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22588         the basic block.  Assert the use reference is not artificial and that
22589         it has an associated insn.
22591 2018-04-04  Michael Matz  <matz@suse.de>
22593         * builtins.c (compute_objsize): Pass correct operand
22594         to array_at_struct_end_p.
22596 2018-04-04  Richard Biener  <rguenther@suse.de>
22598         PR lto/85176
22599         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22600         from contexts for DINFO_LEVEL_TERSE and below.
22602 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22604         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22605         resiter_operand condition.
22606         * config/nds32/nds32.md (*move<mode>): Ditto.
22608 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22609             Monk Chiang  <sh.chiang04@gmail.com>
22611         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22613 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22615         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22617 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22618             Kito Cheng  <kito.cheng@gmail.com>
22620         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22621         nds32_cond_code_str, output_cond_branch,
22622         output_cond_branch_compare_zero, nds32_expand_cbranch,
22623         nds32_expand_cstore, nds32_expand_movcc,
22624         nds32_output_cbranchsi4_equality_zero,
22625         nds32_output_cbranchsi4_equality_reg,
22626         nds32_output_cbranchsi4_equality_reg_or_const_int,
22627         nds32_output_cbranchsi4_greater_less_zero: New functions.
22628         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22629         nds32_expand_cstore, nds32_expand_movcc,
22630         nds32_output_cbranchsi4_equality_zero,
22631         nds32_output_cbranchsi4_equality_reg,
22632         nds32_output_cbranchsi4_equality_reg_or_const_int,
22633         nds32_output_cbranchsi4_greater_less_zero): Declare.
22634         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22635         nds32_rimm11s_operand): New predicates.
22636         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22637         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22638         patterns.
22640 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22642         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22643         * config/nds32/nds32.md: Ditto.
22644         * config/nds32/pipelines.md: Ditto.
22646 2018-04-04  Richard Biener  <rguenther@suse.de>
22648         PR tree-optimization/85168
22649         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22650         propagating abnormals.
22652 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22654         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22656 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22657             Kito Cheng  <kito.cheng@gmail.com>
22659         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22660         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22661         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22662         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22663         * config/nds32/nds32.md (sibcall_internal): New.
22664         (sibcall_register): Remove.
22665         (sibcall_immediate): Remove.
22666         (sibcall_value_internal): New.
22667         (sibcall_value_register): Remove.
22668         (sibcall_value_immediate): Remove.
22669         * config/nds32/predicates.md (nds32_general_register_operand): New.
22670         (nds32_call_address_operand): New.
22672 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22674         PR rtl-optimization/85167
22675         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22676         bb_defs if *split_p, instead preinitialize it to NULL.
22678         PR tree-optimization/85156
22679         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22680         evaluating the argument multiple times.
22682 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22684         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22685         than vector.
22686         (_mm_cvtpd_ps): Likewise.
22687         (_mm_cvttpd_epi32): Likewise.
22688         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22689         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22690         vector, pixel, and bool following altivec.h include.
22692 2018-04-03  Martin Sebor  <msebor@redhat.com>
22694         * doc/extend.texi (Common Function Attributes): Clarify.
22695         (const attribute): Likewise.
22696         (pure attribute): Likewise.
22698 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22700         PR target/85169
22701         * config/i386/i386.c (ix86_expand_vector_set): Use
22702         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22704 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22706         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22707         instructions when changing rounding bits to preserve precision bits
22708         in the x87 control word.
22710 2018-04-03  Martin Liska  <mliska@suse.cz>
22712         PR tree-optimization/82491
22713         * rtl.h (strip_offset_and_add): Replace += suboffset with
22714         poly_uint64 () + suboffset.
22716 2018-03-29  Martin Liska  <mliska@suse.cz>
22717             Martin Jambor  <mjambor@suse.cz>
22719         PR ipa/84947
22720         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22721         param_type is not an integral or pointer type.
22723 2018-04-03  Richard Biener  <rguenther@suse.de>
22725         * sese.h (recompute_all_dominators): Remove.
22727 2018-04-02  Martin Sebor  <msebor@redhat.com>
22729         * doc/invoke.texi (-Wrestrict): Fix typos.
22731 2018-04-02  Jim Wilson  <jimw@sifive.com>
22733         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22734         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22735         (<optab>di3, <optab>si3_extend): Likewise.
22736         (<optab>si3_mask, <optab>si3_mask_1): New.
22737         (<optab>di3_mask, <optab>di3_mask_1): New.
22738         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22739         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22740         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22742 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22744         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22745         example.
22747 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22749         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22750         (nds32_canonicalize_comparison): New function.
22752 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22753             Kito Cheng  <kito.cheng@gmail.com>
22754             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22756         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22757         * config/nds32/constants.md (unspec_volatile_element): Add
22758         UNSPEC_VOLATILE_RELAX_GROUP.
22759         * config/nds32/nds32-relax-opt.c: New file.
22760         * config/nds32/nds32-predicates.c
22761         (nds32_symbol_load_store_p): New function.
22762         * config/nds32/nds32-protos.h
22763         (nds32_symbol_load_store_p): Declare function.
22764         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22765         * config/nds32/nds32.c
22766         (nds32_register_pass): New function to register pass.
22767         (nds32_register_passes): New function to register passes.
22768         * config/nds32/nds32.md (relax_group): New pattern.
22769         * config/nds32/nds32.opt (mrelax-hint): New option.
22770         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22772 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22774         * config/nds32/t-nds32: Modify files dependency.
22776 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22778         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22779         (PROFILE_HOOK): Define its implementation.
22781 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22783         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22784         type and 32-bit size.
22786 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22788         PR middle-end/85090
22789         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22790         (V_128_256): New mode iterator.
22791         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22792         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22793         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22794         of V.
22795         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22796         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22798 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22800         PR target/83315
22801         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22802         NaN inputs correctly.
22804 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22806         PR target/80546
22807         * config/rs6000/vsx.md (??r): New mode attribute.
22808         (*vsx_mov<mode>_64bit): Use it.
22809         (*vsx_mov<mode>_32bit): Likewise.
22811 2018-03-30  Martin Sebor  <msebor@redhat.com>
22813         PR tree-optimization/84818
22814         * builtins.c (check_access): Use warning_n.
22816 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22818         PR target/83822
22819         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22820         condition.
22821         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22822         condition.
22824 2018-03-30  Julia Koval  <julia.koval@intel.com>
22826         PR target/84413
22827         * x86-tune.def (movx, partial_reg_dependency): Enable for
22828         m_SKYLAKE_AVX512.
22830 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22832         PR inline-asm/84985
22833         * lra-constraints.c (process_alt_operands): Move setting
22834         this_alternative_matches below.
22836 2018-03-29  Martin Liska  <mliska@suse.cz>
22838         PR lto/84995.
22839         * doc/invoke.texi: Document how LTO works with debug info.
22840         Describe auto-load support of binutils.  Mention 'x86-64'
22841         as valid option value of -march option.
22843 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22845         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22847         PR c/85094
22848         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22849         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22850         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22851         checking.
22853 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22855         PR target/84912
22856         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22857         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22858         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22859         for RS6000_BTM_POWERPC64.
22860         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22861         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22862         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22863         definition.
22864         (DIVDE): Use it.
22865         (DIVDEU): Likewise.
22867 2018-03-28  Carl Love  <cel@us.ibm.com>
22869         Revert
22870         2017-09-27  Carl Love  <cel@us.ibm.com>
22872         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22873         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22874         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22875         fctiw instruction.
22877 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22879         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22880         instead of __vector bool.
22881         (_mm_max_pu8): Likewise.
22882         (_mm_min_pi16): Likewise.
22884 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22886         PR target/84912
22887         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22888         (DIVWEUO): Likewise.
22889         (DIVDEO): Likewise.
22890         (DIVDEUO): Likewise.
22891         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22892         DIVWEUO and DIVDEUO.
22893         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22894         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22895         (div_extend): Likewise.
22896         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22897         builtin function.
22898         (__builtin_divweuo): Likewise.
22899         (__builtin_divdeo): Likewise.
22900         (__builtin_divdeuo): Likewise.
22902 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22904         PR target/85095
22905         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22906         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22908         PR tree-optimization/82004
22909         * gimple-match-head.c (optimize_pow_to_exp): New function.
22910         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22911         Don't fold to exp if optimize_pow_to_exp is false.
22913 2018-03-28  Martin Liska  <mliska@suse.cz>
22915         PR other/84819
22916         * calls.c (initialize_argument_information): Fix trailing space.
22917         * common.opt: Fix typo and provide better explanation for
22918         -fsanitize-coverage option.
22919         * config/i386/i386.opt: Fix typo.
22921 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
22922             Martin Liska  <mliska@suse.cz>
22924         PR sanitizer/85081
22925         * gimplify.c (asan_poison_variable): Don't do the check for
22926         gimplify_omp_ctxp here.
22927         (gimplify_decl_expr): Do it here.
22928         (gimplify_target_expr): Likewise.
22930 2018-03-28  Martin Liska  <mliska@suse.cz>
22932         PR target/84988
22933         * config/i386/i386.c (ix86_function_arg_advance): Do not call
22934         chkp_type_bounds_count if MPX is not enabled.
22936 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
22938         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22940 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
22942         PR target/84914
22943         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22944         function to create the function decl for complex long double
22945         multiply and divide for -mabi=ieeelongdouble.
22946         (init_float128_ieee): Call it.
22948 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
22950         PR target/85044
22951         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22952         -fcf-protection=branch -mibt.
22953         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22955 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22957         PR target/81863
22958         * config/arm/arm.c (arm_valid_symbolic_address): Handle
22959         arm_word_relocations.
22961 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
22963         PR target/85056
22964         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22965         extern array declarations.
22967 2018-03-27  Richard Biener  <rguenther@suse.de>
22969         PR middle-end/84067
22970         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22971         explicit single_use checks.
22973 2018-03-27  Richard Biener  <rguenther@suse.de>
22975         PR tree-optimization/85082
22976         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22977         Valueize the VUSE.
22979 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22981         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22982         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22983         Turn on fasynchronous-unwind-tables and funwind-tables.
22985 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
22987         PR target/85073
22988         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22989         (*bmi_blsr_<mode>_ccz): Ditto.
22991 2018-03-26  Tom de Vries  <tom@codesourcery.com>
22993         PR tree-optimization/85063
22994         * omp-general.c (offloading_function_p): New function.  Factor out
22995         of ...
22996         * omp-offload.c (pass_omp_target_link::gate): ... here.
22997         * omp-general.h (offloading_function_p): Declare.
22998         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22999         with attribute omp declare target for offloading functions.
23001 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23003         PR tree-optimization/84005
23004         * tree-data-ref.h (get_base_for_alignment): Declare.
23005         * tree-data-ref.c (get_base_for_alignment_1): New function.
23006         (get_base_for_alignment): Likewise.
23007         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23008         get_base_for_alignment to find a suitable base object, instead
23009         of always using drb->base_address.
23011 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23013         PR inline-asm/85022
23014         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23015         known size by default.
23017 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23019         PR inline-asm/85030
23020         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23021         and non BLKmode operands.
23023 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23025         PR target/85026
23026         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23027         Clean up attributes.
23029 2018-03-23  Richard Biener  <rguenther@suse.de>
23031         PR debug/85020
23032         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23033         we are going to emit early debug for LTO.
23035 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23037         PR inline-asm/85034
23038         * function.c (match_asm_constraints_1): Don't optimize if input
23039         doesn't satisfy general_operand predicate for output's mode.
23041         PR inline-asm/85022
23042         * alias.c (write_dependence_p): Don't require for x_canonicalized
23043         non-VOIDmode if x has VOIDmode.
23045         PR sanitizer/85029
23046         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23047         just don't try to optimize it rather than assert it never happens.
23049 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23051         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23052         macro expansions for definition of ST_INTERNAL_<mode> and
23053         LD_INTERNAL_<mode> builtins.
23054         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23055         Remove prototype.
23056         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23057         function.
23058         (altivec_expand_st_builtin): Likewise.
23059         (altivec_expand_builtin): Remove calls to deleted functions.
23060         (rs6000_address_for_altivec): Delete this function.
23061         * config/rs6000/vector.md: Remove expands for
23062         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23064 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23066         PR target/84826
23067         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23068         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23069         re-computing once computed.
23070         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23071         (arm_init_machine_status): Initialize
23072         machine->static_chain_stack_bytes.
23074 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23076         PR target/84760
23077         * doc/extend.texi: Add four new prototypes for vec_ld.
23078         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23079         definitions for more logical presentation.
23080         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23081         entries for V1TI variants of __builtin_altivec_ld builtin.
23082         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23083         handling of V1TI variant of LVX icode pattern.
23084         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23085         (rs6000_gimple_fold_builtin): Likewise.
23086         (altivec_init_builtins): Add code to define
23087         __builtin_altivec_lvx_v1ti function.
23089 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23091         PR inline-asm/84941
23092         * function.c (match_asm_constraints_1): Don't do the optimization
23093         if input isn't a REG, SUBREG, MEM or constant.
23095 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23097         PR tree-optimization/84956
23098         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23099         bb_has_abnormal_pred.
23101 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23103         PR sanitizer/85018
23104         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23105         DECL_INITIAL (decl) to decl at the end.
23106         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23107         adjust the comment.
23109 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23111         * doc/extend.texi (__builtin_tgmath): Document when complex
23112         integer types are treated as _Complex _Float64.
23114 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23116         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23118 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23120         PR tree-optimization/84960
23121         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23122         if it is ENTRY block, move them into single succ of ENTRY in that case.
23124 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23126         PR tree-optimization/84811
23127         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23128         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23129         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23130         (known_subrange_p): Update accordingly.  Cast each value involved
23131         in the size comparison, rather than casting the result of the
23132         subtraction.
23134 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23136         PR tree-optimization/84982
23137         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23138         by flipping the least significant bit rather than all bits from
23139         bitpos to bitpos + bitsize - 1.
23141 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23143         * doc/extend.texi (Deprecated Features): Remove mention of
23144         long-deleted deprecations.
23146 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23148         PR jit/84288
23149         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23150         * configure: Regenerate.
23152 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23154         PR tree-optimization/83126
23155         * tree-parloops.c (num_phis): New function.
23156         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23158 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23160         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23161         mention anon-struct/union members and trailing attributes.
23163 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23165         PR tree-optimization/84969
23166         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23167         builtin memset partitions if they set different rhs values.
23169 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23171         PR rtl-optimization/84989
23172         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23173         VEC_DUPLICATE with scalar result mode.
23175 2018-03-21  Martin Liska  <mliska@suse.cz>
23177         PR ipa/84963
23178         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23179         not intended return statement.
23181 2018-03-21  Martin Liska  <mliska@suse.cz>
23183         PR target/84988
23184         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23185         (chkp_find_bound_slots_1): Limit number of iterations.
23187 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23189         PR target/84838
23190         * Minor grammar fixes for x86 options.
23192 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23194         PR debug/84875
23195         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23196         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23198 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23200         PR target/83789
23201         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23202         (altivec_lvx_<mode>_1op): Likewise.
23203         (altivec_stvx_<mode>_2op): Likewise.
23204         (altivec_stvx_<mode>_1op): Likewise.
23205         (altivec_lvx_<VM2:mode>): New define_expand.
23206         (altivec_stvx_<VM2:mode>): Likewise.
23207         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23208         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23209         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23210         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23211         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23212         (rs6000_gen_lvx): Likewise.
23213         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23214         (altivec_expand_stv_builtin): Likewise.
23215         (altivec_expand_builtin): Likewise.
23216         * config/rs6000/vector.md: Likewise.
23218 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23220         PR target/82518
23221         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23222         BYTES_BIG_ENDIAN.
23224 2018-03-20  Richard Biener  <rguenther@suse.de>
23226         PR target/84986
23227         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23228         sign-conversions as zero, fall back to standard scalar_stmt
23229         cost for the rest.
23231 2018-03-20  Martin Liska  <mliska@suse.cz>
23233         PR ipa/84825
23234         * predict.c (rebuild_frequencies): Handle case when we have
23235         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23237 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23239         PR target/84990
23240         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23241         flag_section_anchors.
23242         * varasm.c (use_blocks_for_decl_p): Remove hack for
23243         dw2_force_const_mem.
23245         PR target/84845
23246         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23247         to ...
23248         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23249         be created, use lowpart_subreg of operands[0] rather than operands[0]
23250         itself.
23251         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23252         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23253         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23254         and n constraint instead of aarch64_shift_imm_di and Usd.
23255         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23256         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23258 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23260         PR target/82989
23261         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23262         to favor GPR over NEON registers.
23263         (<shift>di3_neon): Likewise.
23265 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23267         PR target/84952
23268         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23269         (nvptx_process_pars): Emit bar.sync asap and alap.
23271 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23273         PR target/84954
23274         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23275         seen_label if seen_label is already set.
23277 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23279         PR target/84945
23280         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23281         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23282         Use 1U instead of 1.  Formatting fixes.
23284         PR c/84953
23285         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23286         instead of TREE_TYPE (s1) for the return value.
23288 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23290         PR tree-optimization/84946
23291         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23292         bitsize + bitsize in poly_uint64 rather than poly_int64.
23294         PR sanitizer/78651
23295         * dwarf2asm.c: Include fold-const.c.
23296         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23297         of decl rather than decl itself.
23299         PR rtl-optimization/84643
23300         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23302 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23304         PR sanitizer/78651
23305         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23306         calling assemble_variable.
23308 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23310         PR target/81647
23311         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23312         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23314 2018-03-19  Jim Wilson  <jimw@sifive.com>
23316         PR bootstrap/84856
23317         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23318         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23319         (riscv_first_stack_step): Likewise.
23320         (riscv_option_override): Use STACK_BOUNDARY instead of
23321         MIN_STACK_BOUNDARY.
23322         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23323         MIN_STACK_BOUNDARY.
23324         (BIGGEST_ALIGNMENT): Set to 128.
23325         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23326         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23327         STACK_BOUNDARY.
23329 2018-03-19  Richard Biener  <rguenther@suse.de>
23331         PR tree-optimization/84933
23332         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23333         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23335 2018-03-19  Richard Biener  <rguenther@suse.de>
23337         PR tree-optimization/84859
23338         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23339         (cond_if_else_store_replacement): Perform sinking operation on
23340         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23341         Generalize what a BB with a single eligible store is.
23343 2018-03-19  Richard Biener  <rguenther@suse.de>
23345         PR tree-optimization/84929
23346         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23347         chrec_is_positive against non-chrec arg.
23349 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23351         PR target/84711
23352         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23354 2018-03-18  Martin Liska  <mliska@suse.cz>
23356         PR rtl-optimization/84635
23357         * regrename.c (build_def_use): Use matches_mode only when
23358         matches >= 0.
23360 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23362         PR tree-optimization/84913
23363         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23364         vectorize chains of COND_EXPRs.
23366 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23368         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23370 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23372         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23374 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23376         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23378 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23379             Kito Cheng  <kito.cheng@gmail.com>
23381         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23382         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23383         (nds32_adjust_reg_alloc_order): New function.
23384         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23386 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23388         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23389         nds32_print_operand, nds32_print_operand_address): Use
23390         HOST_WIDE_INT_PRINT_DEC instead.
23392 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23394         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23396 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23398         PR target/84902
23399         * config/i386/i386.c (initial_ix86_tune_features,
23400         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23401         unsigned long long.
23402         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23403         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23404         rather than 1u << ix86_tune.  Formatting fix.
23405         (ix86_option_override_internal): Change ix86_arch_mask from
23406         unsigned int to unsigned HOST_WIDE_INT, initialize to
23407         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23408         (ix86_function_specific_restore): Likewise.
23410 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23412         PR target/84899
23413         * postreload.c (reload_combine_recognize_pattern): Perform
23414         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23415         truncate_int_for_mode the result for the destination's mode.
23417         PR c/84909
23418         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23419         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23420         Likewise.
23421         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23422         Likewise.
23424 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23426         PR target/84876
23427         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23428         regno_allocno_class_array and sorted_pseudos.
23429         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23430         insns where regno is used.
23432 2018-03-16  Martin Liska  <mliska@suse.cz>
23434         PR ipa/84833
23435         * multiple_target.c (create_dispatcher_calls): Redirect
23436         reference in the symbol table.
23438 2018-03-16  Martin Liska  <mliska@suse.cz>
23440         PR ipa/84722
23441         * multiple_target.c (create_dispatcher_calls): Redirect also
23442         an alias.
23444 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23446         PR c++/79937
23447         PR c++/82410
23448         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23449         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23450         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23452 2018-03-16  Julia Koval  <julia.koval@intel.com>
23454         * doc/invoke.texi (Skylake Server): Add CLWB.
23455         Cannonlake): Remove CLWB.
23457 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23459         PR tree-optimization/84841
23460         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23461         1 << 3.
23462         (FLOAT_ONE_CONST_TYPE): Define.
23463         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23464         (sort_by_operand_rank): Put entries with higher constant_type last
23465         rather than first to match comments.
23467 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23469         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23470         split predicate.
23472 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23474         PR c++/79085
23475         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23476         check and use address of target always.
23478 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23480         PR target/84574
23481         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23482         (indirect_thunk_bnd_needed): Likewise.
23483         (indirect_thunks_used): Likewise.
23484         (indirect_thunks_bnd_used): Likewise.
23485         (indirect_return_needed): New.
23486         (indirect_return_bnd_needed): Likewise.
23487         (output_indirect_thunk_function): Add a bool argument for
23488         function return.
23489         (output_indirect_thunk_function): Don't generate alias for
23490         function return thunk.
23491         (ix86_code_end): Call output_indirect_thunk_function to generate
23492         function return thunks.
23493         (ix86_output_function_return): Set indirect_return_bnd_needed
23494         and indirect_return_needed instead of indirect_thunk_bnd_needed
23495         and indirect_thunk_needed.
23497 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23499         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23500         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23501         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23503 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23504             Paul Hua <paul.hua.gm@gmail.com>
23506         PR c/84852
23507         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23509 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23511         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23512         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23513         resp. SFmode cases.
23515 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23517         PR target/84711
23518         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23519         instead of GET_MODE_SIZE when comparing Units.
23521 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23523         PR target/68256
23524         * varasm.c (hash_section): Return an unchangeble hash value
23525         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23526         Return !aarch64_can_use_per_function_literal_pools_p ().
23528 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23530         PR target/84860
23531         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23532         rather than address of cmode as last argument to prepare_cmp_insn.
23534 2018-03-15  Julia Koval  <julia.koval@intel.com>
23536         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23537         F_AVX512VNNI, F_AVX512BITALG): New.
23539 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23541         PR target/83451
23542         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23543         insn for floating-point loads and stores.
23545 2018-03-14  Carl Love  <cel@us.ibm.com>
23547         * config/rs6000/rs6000-c.c: Add macro definitions for
23548         ALTIVEC_BUILTIN_VEC_PERMXOR.
23549         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23550         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23551         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23552         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23553         UNSPEC_VPERMXOR.
23554         * config/doc/extend.texi: Add prototypes for vec_permxor.
23556 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23558         PR c/84852
23559         * diagnostic-show-locus.c (class layout_point): Convert m_line
23560         from int to linenum_type.
23561         (line_span::comparator): Use linenum "compare" function when
23562         comparing line numbers.
23563         (test_line_span): New function.
23564         (layout_range::contains_point): Convert param "row" from int to
23565         linenum_type.
23566         (layout_range::intersects_line_p): Likewise.
23567         (layout::will_show_line_p): Likewise.
23568         (layout::print_source_line): Likewise.
23569         (layout::should_print_annotation_line_p): Likewise.
23570         (layout::print_annotation_line): Likewise.
23571         (layout::print_leading_fixits): Likewise.
23572         (layout::annotation_line_showed_range_p): Likewise.
23573         (struct line_corrections): Likewise for field m_row.
23574         (line_corrections::line_corrections): Likewise for param "row".
23575         (layout::print_trailing_fixits): Likewise.
23576         (layout::get_state_at_point): Likewise.
23577         (layout::get_x_bound_for_row): Likewise.
23578         (layout::print_line): Likewise.
23579         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23580         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23581         * input.c (selftest::test_linenum_comparisons): New function.
23582         (selftest::input_c_tests): Call it.
23583         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23584         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23585         * selftest.h (ASSERT_GT): New macro.
23586         (ASSERT_GT_AT): New macro.
23587         (ASSERT_LT): New macro.
23588         (ASSERT_LT_AT): New macro.
23590 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23592         PR rtl-optimization/84780
23593         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23595 2018-03-14  Martin Liska  <mliska@suse.cz>
23597         * tree.c (record_node_allocation_statistics): Use
23598         get_stats_node_kind.
23599         (get_stats_node_kind): New function extracted from
23600         record_node_allocation_statistics.
23601         (free_node): Use get_stats_node_kind.
23603 2018-03-14  Richard Biener  <rguenther@suse.de>
23605         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23606         that the value-set of ANTIC_IN doesn't grow.
23608         Revert
23609         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23610         member.
23611         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23612         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23614 2018-03-14  Julia Koval  <julia.koval@intel.com>
23616         * config.gcc (icelake-client, icelake-server): New.
23617         (icelake): Remove.
23618         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23619         (initial_ix86_arch_features): Ditto.
23620         (PTA_SKYLAKE): Add SGX.
23621         (PTA_ICELAKE): Remove.
23622         (PTA_ICELAKE_CLIENT): New.
23623         (PTA_ICELAKE_SERVER): New.
23624         (ix86_option_override_internal): Split up icelake on icelake client and
23625         icelake server.
23626         (get_builtin_code_for_version): Ditto.
23627         (fold_builtin_cpu): Ditto.
23628         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23629         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23630         * config/i386/i386.h (processor_type): Ditto.
23631         * doc/invoke.texi: Ditto.
23633 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23635         PR sanitizer/83392
23636         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23637         INTEGER_CST offset, add it together with bitpos / 8 and
23638         sign extend based on POINTER_SIZE.
23640         PR target/84844
23641         Revert
23642         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23644         PR target/78090
23645         * config/i386/constraints.md (Yc): New register constraint.
23646         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23647         Use Yc constraint for alternative 2 of operand 0.  Remove
23648         preferred_for_speed attribute.
23650 2018-03-14  Richard Biener  <rguenther@suse.de>
23652         PR tree-optimization/84830
23653         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23654         with the old one to avoid oscillations.
23656 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23658         PR target/83712
23659         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23660         pseudos.
23661         (assign_by_spills): Return a flag of reload assignment failure.
23662         Do not process the reload assignment failures.  Do not spill other
23663         reload pseudos if they has the same reg class.  Update n if
23664         necessary.
23665         (lra_assign): Add a return arg.  Set up from the result of
23666         assign_by_spills call.
23667         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23668         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23669         usage_insns if it is not NULL.
23670         (spill_hard_reg_in_range): New function.
23671         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23672         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23673         function prototypes.
23674         (lra_assign): Change prototype.
23675         * lra.c (lra): Add code to deal with fails by splitting hard reg
23676         live ranges.
23678 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23680         * config/riscv/riscv.opt (mrelax): New option.
23681         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23682         "norelax" when riscv_mrelax is disabled.
23683         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23685 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23687         PR target/84743
23688         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23689         reassociation for int modes.
23691 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23693         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23694         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23695         for big-endian.
23696         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23697         * config/aarch64/aarch64-sve.md
23698         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23699         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23700         (*extend<mode><Vwide>2): Rename to...
23701         (aarch64_sve_extend<mode><Vwide>2): ...this.
23702         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23703         renaming the old pattern to...
23704         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23705         unsigned packs.
23706         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23707         define_expand, renaming the old pattern to...
23708         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23709         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23710         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23711         account when deciding which SVE instruction the optab should use.
23712         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23714 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23716         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23717         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23718         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23719         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23720         (tlsdesc_small_<mode>): Turn a define_expand and use
23721         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23722         (tlsdesc_small_advsimd_<mode>): ...this.
23723         (tlsdesc_small_sve_<mode>): New pattern.
23725 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23727         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23728         (UNSPEC_UMUL_HIGHPART): New constants.
23729         (MUL_HIGHPART): New int iteraor.
23730         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23731         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23732         define_expand.
23733         (*<su>mul<mode>3_highpart): New define_insn.
23735 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23737         PR lto/84805
23738         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23739         incomplete types.
23741 2018-03-13  Martin Liska  <mliska@suse.cz>
23743         PR ipa/84658.
23744         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23745         vector.
23746         (sem_item_optimizer::~sem_item_optimizer): Release it.
23747         (sem_item_optimizer::merge_classes): Register variable aliases.
23748         (sem_item_optimizer::fixup_pt_set): New function.
23749         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23750         * ipa-icf.h: Declare new variables and functions.
23752 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23754         PR middle-end/84834
23755         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23756         integer_pow2p@2 and test integer_pow2p in condition.
23757         (A < 0 ? C : 0): Similarly for @1.
23759         PR middle-end/84831
23760         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23761         characters starting at p contain '\0' character, don't look beyond
23762         that.
23764         PR target/84827
23765         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23766         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23768         PR target/84828
23769         * reg-stack.c (change_stack): Change update_end var from int to
23770         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23771         also call set_block_for_insn on the newly added insns and rescan.
23773         PR target/84786
23774         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23775         on the last operand.
23777         PR c++/84704
23778         * tree.c (stabilize_reference_1): Return save_expr (e) for
23779         STATEMENT_LIST even if it doesn't have side-effects.
23781 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23783         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23785 2018-03-12  Renlin Li  <renlin.li@arm.com>
23787         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23788         aarch64_output_scalar_simd_mov_immediate.
23790 2018-03-12  Martin Sebor  <msebor@redhat.com>
23792         PR tree-optimization/83456
23793         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23794         for perfectly overlapping calls to memcpy.
23795         (gimple_fold_builtin_memory_chk): Same.
23796         (gimple_fold_builtin_strcpy): Handle no-warning.
23797         (gimple_fold_builtin_stxcpy_chk): Same.
23798         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23800 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23802         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23803         parameter.  Use it for SFmode.
23804         (rs6000_function_arg_advance_1): Adjust.
23805         (rs6000_function_arg): Adjust.
23806         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23808 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23810         PR rtl-optimization/84169
23811         PR rtl-optimization/84780
23812         * combine.c (can_combine_p): Check for a 2-insn combination whether
23813         the destination register is used between the two insns, too.
23815 2018-03-12  Richard Biener  <rguenther@suse.de>
23817         PR tree-optimization/84803
23818         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23819         for refs DR analysis didn't process.
23821 2018-03-12  Richard Biener  <rguenther@suse.de>
23823         PR tree-optimization/84777
23824         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23825         force-vectorize loops ignore whether we are optimizing for size.
23827 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23829         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23830         (TARGET_MD_ASM_ADJUST): Define.
23832 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23833             Kito Cheng  <kito.cheng@gmail.com>
23834             Chung-Ju Wu  <jasonwucj@gmail.com>
23836         * config/nds32/nds32.c (nds32_compute_stack_frame,
23837         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23838         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23839         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23840         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23841         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23842         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23843         * config/nds32/nds32.md (prologue, epilogue): Use macro
23844         NDS32_V3PUSH_AVAILABLE_P to do checking.
23846 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23848         PR debug/58150
23849         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23850         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23851         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23852         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23853         addition of most attributes on !orig_type_die or the attribute not
23854         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23856 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23857             Chung-Ju Wu  <jasonwucj@gmail.com>
23859         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23860         __NDS32_VH__ macro.
23861         * config/nds32/nds32.opt (mvh): New option.
23863 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23864             Chung-Ju Wu  <jasonwucj@gmail.com>
23866         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23867         function.
23868         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23869         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23870         definition.
23872 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23873             Chung-Ju Wu  <jasonwucj@gmail.com>
23875         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23876         function.
23877         * config/nds32/nds32-multiple.md (strlensi): New pattern.
23878         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23880 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
23881             Kito Cheng  <kito.cheng@gmail.com>
23882             Chung-Ju Wu  <jasonwucj@gmail.com>
23884         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23885         UNSPEC_FFMISM and UNSPEC_FLMISM.
23886         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23887         for ffb, ffmism and flmism.
23888         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23889         (unspec_ffmism): Ditto.
23890         (unspec_flmism): Ditto.
23891         (nds32_expand_builtin_impl): Check if string extension is available.
23892         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23893         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23895 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
23897         Reverting patch:
23898         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23900         PR target/83712
23901         * lra-assigns.c (assign_by_spills): Return a flag of reload
23902         assignment failure.  Do not process the reload assignment
23903         failures.  Do not spill other reload pseudos if they has the same
23904         reg class.
23905         (lra_assign): Add a return arg.  Set up from the result of
23906         assign_by_spills call.
23907         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23908         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23909         usage_insns if it is not NULL.
23910         (spill_hard_reg_in_range): New function.
23911         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23912         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23913         function prototypes.
23914         (lra_assign): Change prototype.
23915         * lra.c (lra): Add code to deal with fails by splitting hard reg
23916         live ranges.
23918 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
23920         PR target/84807
23921         * config/i386/i386.opt: Replace Enforcment with Enforcement.
23923 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
23925         PR debug/84620
23926         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23927         (dw_val_node): Add val_symbolic_view.
23928         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23929         (symview_upper_bound): New.
23930         (new_line_info_table): Initialize symviews_since_reset.
23931         (dwarf2out_source_line): Count symviews_since_reset and set
23932         symview_upper_bound.
23933         (dw_val_equal_p): Handle symview.
23934         (add_AT_symview): New.
23935         (print_dw_val): Handle symview.
23936         (attr_checksum, attr_checksum_ordered): Likewise.
23937         (same_dw_val_p, size_of_die): Likewise.
23938         (value_format, output_die): Likewise.
23939         (add_high_low_attributes): Use add_AT_symview for entry_view.
23940         (dwarf2out_finish): Reset symview_upper_bound, clear
23941         zero_view_p.
23943 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
23945         PR target/83969
23946         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23947         Add strict argument and use it.
23948         (rs6000_split_multireg_move): Update for new strict argument.
23949         (mem_operand_gpr): Disallow all non-offsettable addresses.
23950         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23952 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
23954         PR target/84772
23955         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23956         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23957         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23959         PR c++/84767
23960         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23961         decl, use remap_type if we want to use the type.
23963 2018-03-09  Martin Sebor  <msebor@redhat.com>
23965         PR tree-optimization/84526
23966         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23967         Remove dead code.
23968         (builtin_access::generic_overlap): Be prepared to handle non-array
23969         base objects.
23971 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
23973         PR rtl-optimization/84682
23974         * lra-constraints.c (process_address_1): Check is_address flag
23975         for address constraints.
23976         (process_alt_operands): Likewise.
23977         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23978         preprocess_constraints.
23979         * recog.h (preprocess_constraints): Add oploc parameter.
23980         Adjust callers.
23981         * recog.c (preprocess_constraints): Test address_operand for
23982         CT_ADDRESS constraints.
23984 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
23986         PR target/83712
23987         * lra-assigns.c (assign_by_spills): Return a flag of reload
23988         assignment failure.  Do not process the reload assignment
23989         failures.  Do not spill other reload pseudos if they has the same
23990         reg class.
23991         (lra_assign): Add a return arg.  Set up from the result of
23992         assign_by_spills call.
23993         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23994         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23995         usage_insns if it is not NULL.
23996         (spill_hard_reg_in_range): New function.
23997         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23998         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23999         function prototypes.
24000         (lra_assign): Change prototype.
24001         * lra.c (lra): Add code to deal with fails by splitting hard reg
24002         live ranges.
24004 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24006         PR target/83193
24007         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24008         Accept complain bool parameter.  Only emit errors if it is true.
24009         (arm_parse_cpu_option_name): Likewise.
24010         (arm_target_thumb_only): Adjust callers of the above.
24011         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24012         prototype to take a default true bool parameter.
24013         (arm_parse_arch_option_name): Likewise.
24015 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24016             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24018         PR jit/64089
24019         PR jit/84288
24020         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24021         * configure: Regenerate.
24022         * configure.ac ("linker --version-script option"): New.
24023         ("linker soname option"): New.
24025 2018-03-09  Richard Biener  <rguenther@suse.de>
24027         PR tree-optimization/84775
24028         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24029         immediate uses of predicate stmts and mark them modified.
24031         Revert
24032         PR tree-optimization/84178
24033         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24034         to caller.
24035         (version_loop_for_if_conversion): Delay update_ssa call.
24036         (tree_if_conversion): Delay update_ssa until after predicate
24037         insertion.
24039 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24041         PR target/84763
24042         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24043         when the function accesses prior frames.
24045 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24047         PR debug/84456
24048         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24049         gen_llsym, otherwise call maybe_gen_llsym.
24051         PR inline-asm/84742
24052         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24053         has ',' character inside of it.
24055 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24057         PR target/84748
24058         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24059         as clobbering CC_REGNUM.
24061 2018-03-08  Richard Biener  <rguenther@suse.de>
24063         PR middle-end/84552
24064         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24065         (follow_copies_to_constant): Do not follow SSA names registered
24066         for update.
24068 2018-03-08  Richard Biener  <rguenther@suse.de>
24070         PR tree-optimization/84178
24071         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24072         to caller.
24073         (version_loop_for_if_conversion): Delay update_ssa call.
24074         (tree_if_conversion): Delay update_ssa until after predicate
24075         insertion.
24077 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24079         PR tree-optimization/84178
24080         * tree-if-conv.c (release_bb_predicate): Remove the
24081         the assertion that the stmts have NULL use_ops.
24082         Discard the statements, asserting that they haven't
24083         yet been added to a BB.
24085 2018-03-08  Richard Biener  <rguenther@suse.de>
24087         PR tree-optimization/84746
24088         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24089         (phi_translate): Pass in destination ANTIC_OUT set.
24090         (phi_translate_1): Likewise.  For a simplified result lookup
24091         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24092         (phi_translate_set): Adjust.
24093         (do_pre_regular_insertion): Likewise.
24094         (do_pre_partial_partial_insertion): Likewise.
24096 2018-03-08  Martin Liska  <mliska@suse.cz>
24098         PR gcov-profile/84735
24099         * doc/gcov.texi: Document usage of profile files.
24100         * gcov-io.h: Document changes in the format.
24102 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24104         PR debug/84404
24105         PR debug/84408
24106         * dwarf2out.c (struct dw_line_info_table): Update comments for
24107         view == -1.
24108         (FORCE_RESET_NEXT_VIEW): New.
24109         (FORCE_RESETTING_VIEW_P): New.
24110         (RESETTING_VIEW_P): Check for -1 too.
24111         (ZERO_VIEW_P): Likewise.
24112         (new_line_info_table): Force-reset next view.
24113         (dwarf2out_begin_function): Likewise.
24114         (dwarf2out_source_line): Simplify zero_view_p initialization.
24115         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24116         view directly.  Omit view when omitting .loc at line 0.
24118 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24120         PR tree-optimization/84740
24121         * tree-switch-conversion.c (process_switch): Call build_constructors
24122         only if info.phi_count is non-zero.
24124         PR tree-optimization/84739
24125         * tree-tailcall.c (find_tail_calls): Check call arguments against
24126         DECL_ARGUMENTS (current_function_decl) rather than
24127         DECL_ARGUMENTS (func) when checking for tail recursion.
24129 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24131         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24132         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24133         Volker Reichelt's entry and add entries for people that perform
24134         GCC fuzzy testing and report numerous bugs.
24136 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24138         PR target/82411
24139         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24140         readonly data in sdata, if that is disabled.
24141         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24142         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24143         -mreadonly-in-sdata option.
24145 2018-03-07  Martin Sebor  <msebor@redhat.com>
24147         PR tree-optimization/84468
24148         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24149         basic block when looking for nul assignment.
24151 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24153         PR target/84277
24154         * except.h (output_function_exception_table): Adjust prototype.
24155         * except.c (output_function_exception_table): Remove FNNAME parameter
24156         and add SECTION parameter.  Ouput one part of the table at a time.
24157         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24158         the first part of the exception table and emit unwind directives.
24159         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24160         (i386_pe_seh_cold_init): Likewise.
24161         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24162         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24163         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24164         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24165         * config/i386/winnt.c: Include except.h.
24166         (struct seh_frame_state): Add reg_offset, after_prologue and
24167         in_cold_section fields.
24168         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24169         (i386_pe_seh_cold_init): New function.
24170         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24171         to seh->in_cold_section.
24172         (seh_emit_push): Record the offset of the push.
24173         (seh_emit_save): Record the offet of the save.
24174         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24175         Test seh->after_prologue to disregard the epilogue.
24176         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24177         (i386_pe_end_cold_function): New function.
24179 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24181         PR fortran/84565
24182         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24183         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24185         PR c++/84704
24186         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24187         on tmp_var.
24188         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24189         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24191         PR middle-end/84723
24192         * multiple_target.c: Include tree-inline.h and intl.h.
24193         (expand_target_clones): Diagnose and fail if node->definition and
24194         !tree_versionable_function_p (node->decl).
24196 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24198         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24199         sprint_ul.
24200         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24201         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24202         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24204 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24206         PR target/84710
24207         * combine.c (try_combine): Use reg_or_subregno instead of handling
24208         just paradoxical SUBREGs and REGs.
24210 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24212         * config/arc/arc.c (arc_finalize_pic): Remove function.
24213         (arc_must_save_register): We use single base PIC register, remove
24214         checks to save/restore the PIC register.
24215         (arc_expand_prologue): Likewise.
24216         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24217         Remove.
24218         (arc_verify_short): Likewise.
24219         (arc_attr_type): Likewise.
24220         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24221         (walk_stores): Likewise.
24222         (arc_address_cost): Make it static.
24223         (arc_verify_short): Likewise.
24224         (branch_dest): Likewise.
24225         (arc_attr_type): Likewise.
24226         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24227         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24228         (arc_final_prescan_insn): Remove inserting the nops due to
24229         hardware hazards.  It is done in reorg step.
24230         (insn_length_variant_t): Remove.
24231         (insn_length_parameters_t): Likewise.
24232         (arc_insn_length_parameters): Likewise.
24233         (arc_get_insn_variants): Likewise.
24234         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24236 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24238         PR inline-asm/84683
24239         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24240         assertion failure.
24242         PR tree-optimization/84687
24243         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24244         on new_node->decl.
24245         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24247 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24249         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24250         Rename to ppc_speculation_barrier.
24251         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24252         __builtin_ppc_speculation_barrier.
24254 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24256         PR target/84700
24257         * combine.c (combine_simplify_rtx): Don't try to simplify if
24258         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24259         are equal to x.
24261 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24263         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24264         to 32 bytes when compiling for POWER9.
24266 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24268         PR target/84564
24269         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24270         regparm >= 3 with no arg reg available also for calls with
24271         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24273         PR target/84524
24274         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24275         orig,vex.
24276         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24278 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24280         PR target/84264
24281         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24283 2018-03-05  Richard Biener  <rguenther@suse.de>
24285         PR tree-optimization/84486
24286         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24287         When inserting a __builtin_assume_aligned call set the LHS
24288         SSA name alignment info accordingly.
24290 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24292         PR tree-optimization/84114
24293         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24294         Avoid reassociation of FLOAT_MODE addition.
24296 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24298         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24299         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24300         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24301         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24302         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24303         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24304         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24305         and -mwbnoinvd.
24306         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24307         __builtin_ia32_wbinvd): New builtins.
24308         (SPECIAL_ARGS2): New.
24309         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24310         (SPECIAL_ARGS2): New.
24311         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24312         (ix86_valid_target_attribute_inner_p): Ditto.
24313         (ix86_init_mmx_sse_builtins): Add special_args2.
24314         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24315         TARGET_WBNOINVD_P): New.
24316         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24317         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24318         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24319         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24320         * config/i386/pconfigintrin.h: New file.
24321         * config/i386/wbnoinvdintrin.h: Ditto.
24322         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24323         wbnoinvdintrin.h.
24324         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24326 2018-03-05  Richard Biener  <rguenther@suse.de>
24328         PR tree-optimization/84670
24329         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24330         member.
24331         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24332         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24333         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24334         doesn't grow if all successors (recursively) were visited at least
24335         once.
24337 2018-03-05  Richard Biener  <rguenther@suse.de>
24339         PR tree-optimization/84650
24340         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24341         if executed in the loop pipeline.
24343 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24345         * doc/configfiles.texi (Configuration Files): Move info about
24346         conditionalizing $target-protos.h to...
24347         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24348         differs from $target-protos.h.
24350 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24351             Chung-Ju Wu  <jasonwucj@gmail.com>
24353         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24354         * config/nds32/nds32-multiple.md (setmemsi): Define.
24355         * config/nds32/nds32-memory-manipulation.c
24356         (nds32_gen_dup_4_byte_to_word_value): New.
24357         (emit_setmem_word_loop): New.
24358         (emit_setmem_byte_loop): New.
24359         (nds32_expand_setmem_loop): New.
24360         (nds32_expand_setmem_loop_v3m): New.
24361         (nds32_expand_setmem_unroll): New.
24362         (nds32_expand_setmem): New.
24364 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24365             Chung-Ju Wu  <jasonwucj@gmail.com>
24367         * config/nds32/nds32-memory-manipulation.c
24368         (nds32_emit_load_store): New.
24369         (nds32_emit_post_inc_load_store): New.
24370         (nds32_emit_mem_move): New.
24371         (nds32_emit_mem_move_block): New.
24372         (nds32_expand_movmemsi_loop_unknown_size): New.
24373         (nds32_expand_movmemsi_loop_known_size): New.
24374         (nds32_expand_movmemsi_loop): New.
24375         (nds32_expand_movmemsi_unroll): New.
24376         (nds32_expand_movmemqi): Rename ...
24377         (nds32_expand_movmemsi): ... to this.
24378         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24379         (movmemsi): ... to this.
24380         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24381         (nds32_expand_movmemsi): ... to this.
24383 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24384             Monk Chiang  <sh.chiang04@gmail.com>
24385             Chung-Ju Wu  <jasonwucj@gmail.com>
24387         * config/nds32/nds32-protos.h
24388         (nds32_expand_load_multiple): New arguments.
24389         (nds32_expand_store_multiple): Ditto.
24390         (nds32_valid_multiple_load_store): Rename ...
24391         (nds32_valid_multiple_load_store_p): ... to this.
24392         * config/nds32/nds32-memory-manipulation.c
24393         (nds32_expand_load_multiple): Refine implementation.
24394         (nds32_expand_store_multiple): Ditto.
24395         * config/nds32/nds32-multiple.md
24396         (load_multiple): Update nds32_expand_load_multiple interface.
24397         (store_multiple): Update nds32_expand_store_multiple interface.
24398         * config/nds32/nds32-predicates.c
24399         (nds32_valid_multiple_load_store): Rename ...
24400         (nds32_valid_multiple_load_store_p): ... to this and refine
24401         implementation.
24402         * config/nds32/predicates.md
24403         (nds32_load_multiple_and_update_address_operation): New predicate.
24404         (nds32_store_multiple_and_update_address_operation): New predicate.
24406 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24407             Chung-Ju Wu  <jasonwucj@gmail.com>
24409         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24410         (combo): New attribute.
24411         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24413 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24415         * config/nds32/nds32.opt: Change -mcmodel= default value.
24417 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24418             Monk Chiang  <sh.chiang04@gmail.com>
24419             Chung-Ju Wu  <jasonwucj@gmail.com>
24421         * config/nds32/constants.md (unspec_element): New enum.
24422         * config/nds32/constraints.md (Umw): New constraint.
24423         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24424         * config/nds32/nds32-intrinsic.md: Likewise.
24425         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24426         (nds32_valid_smw_lwm_base_p): New.
24427         (nds32_output_smw_single_word): New.
24428         (nds32_output_lmw_single_word): New.
24429         (nds32_expand_unaligned_load): New.
24430         (nds32_expand_unaligned_store): New.
24431         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24432         (nds32_output_smw_single_word): Declare.
24433         (nds32_output_lmw_single_word): Declare.
24434         (nds32_expand_unaligned_load): Declare.
24435         (nds32_expand_unaligned_store): Declare.
24436         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24437         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24438         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24439         NDS32_BUILTIN_UASTORE_DW.
24440         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24441         predicate.
24443 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24444             Kito Cheng  <kito.cheng@gmail.com>
24445             Chung-Ju Wu  <jasonwucj@gmail.com>
24447         * config/nds32/nds32-intrinsic.c
24448         (nds32_expand_builtin_null_ftype_reg): Delete.
24449         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24450         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24451         (nds32_read_argument): New.
24452         (nds32_legitimize_target): Ditto.
24453         (nds32_legitimize_argument): Ditto.
24454         (nds32_check_constant_argument): Ditto.
24455         (nds32_expand_unop_builtin): Ditto.
24456         (nds32_expand_unopimm_builtin): Ditto.
24457         (nds32_expand_binop_builtin): Ditto.
24458         (nds32_builtin_decl_impl): Ditto.
24459         (builtin_description): Ditto.
24460         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24461         (nds32_init_builtins_impl): Ditto.
24462         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24463         (nds32_builtin_decl): New.
24464         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24465         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24467 2018-03-02  Jeff Law  <law@redhat.com>
24469         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24470         (redundant_insn, fill_simple_delay_slots): Likewise.
24471         (fill_slots_from_thread): Likewise.
24472         * resource.c (mark_referenced_resources): Likewise.
24473         (mark_set_resources, find_dead_or_set_registers): Likewise.
24475 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24477         * substring-locations.h (format_warning_va): Formatting fix for
24478         ATTRIBUTE_GCC_DIAG.
24479         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24480         argument.
24481         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24482         * substring-locations.c: Include intl.h.
24483         (format_warning_va): Turned into small wrapper around
24484         format_warning_n_va, renamed to ...
24485         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24486         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24487         use ngettext.
24488         (format_warning_at_substring_n): New function.
24489         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24490         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24491         format_warning_at_substring with just a shorter name instead of
24492         const function pointer.
24493         (fmtwarn_n): New function.
24494         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24495         appropriate, get rid of all the fmtstr temporaries, move conditionals
24496         with G_() wrapped string literals directly into fmtwarn arguments,
24497         cast dir.len to (int), formatting fixes.
24499 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24501         * doc/invoke.texi: Remove "Cilk Plus" references.
24503 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24504             Richard Biener  <rguenther@suse.de>
24506         PR ipa/84628
24507         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24508         for error or warning attributes if CALL_FROM_THUNK_P is set.
24509         Formatting fixes.
24511 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24513         PR target/56540
24514         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24515         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24517         PR target/56540
24518         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24519         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24521         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24522         instead of -1U in last predictors element's probability member.
24524 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24526         PR ipa/83983
24527         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24528         arguments if they are comparable.
24530 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24532         PR tree-optimization/84634
24533         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24534         masks and masked_loop_p with a single loop_masks, making sure it's
24535         null for bb vectorization.
24537 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24539         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24540         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24541         loop->force_vectorize to check whether there is no alias.
24543 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24545         PR target/84614
24546         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24547         prototypes.
24548         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24549         comments.
24550         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24551         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24552         instead of a loop around prev_real_insn.
24553         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24554         prev_real_insn.
24556         PR inline-asm/84625
24557         * config/i386/i386.c (ix86_print_operand): Use conditional
24558         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24559         zero vector.
24561 2018-03-02  Richard Biener  <rguenther@suse.de>
24563         PR tree-optimization/84427
24564         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24565         (bitmap_set_subtract_values): Rewrite to handle multiple
24566         exprs per value.
24567         (clean): Likewise.
24568         (prune_clobbered_mems): Likewise.
24569         (phi_translate): Take edge instead of pred/phiblock.
24570         (phi_translate_1): Likewise.
24571         (phi_translate_set): Likewise.  Insert all translated
24572         exprs for a value into the set, keeping possibly multiple
24573         expressions per value.
24574         (compute_antic_aux): Adjust for phi_translate changes.
24575         When intersecting union the expressions and prune those
24576         not in the final value set, keeping possibly multiple
24577         expressions per value.  Do not use value-insertion
24578         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24579         all expressions.  Add verification that the value-sets
24580         only shrink during iteration.
24581         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24582         (do_pre_regular_insertion): Likewise.
24583         (do_pre_partial_partial_insertion): Likewise.
24585 2018-03-02  Richard Biener  <rguenther@suse.de>
24587         PR target/82005
24588         * config/darwin.c (saved_debug_info_level): New static global.
24589         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24590         (darwin_asm_lto_end): Restore debug info generation settings.
24592 2018-03-01  Martin Liska  <mliska@suse.cz>
24594         PR sanitizer/82484
24595         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24596         volatile arguments.
24598 2018-03-01  Richard Biener  <rguenther@suse.de>
24600         PR debug/84645
24601         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24602         type annotation with LTO when debug was disabled at compile-time.
24604 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24606         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24607         XINT with INTVAL.
24608         (mips_final_postscan_insn): Likewise.
24610 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24612         PR rtl-optimization/84528
24613         * alias.c (init_alias_target): Add commentary.
24614         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24615         a unique base value if the frame pointer is not eliminated
24616         to the stack pointer.
24618 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24620         PR rtl-optimization/83327
24621         * lra-int.h (hard_regs_spilled_into): Declare.
24622         * lra.c (hard_regs_spilled_into): Define.
24623         (init_reg_info): Init hard_regs_spilled_into.
24624         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24625         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24626         (process_bb_lives): Handle hard_regs_spilled_into.
24627         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24628         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24630 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24632         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24633         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24634         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24635         * config/rs6000/aix72.h: New file.
24637 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24639         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24640         instead of warning_at with conditional singular and plural messages
24641         where possible.
24643         PR target/52991
24644         * stor-layout.c (update_alignment_for_field): For
24645         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24646         && !DECL_PACKED (field), do the alignment update, just use
24647         only desired_align instead of MAX (type_align, desired_align)
24648         as the alignment.
24649         (place_field): Don't do known_align < desired_align handling
24650         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24651         is non-NULL, instead do it after rli->prev_field handling and
24652         only if not within a bitfield word.  For DECL_PACKED (field)
24653         use type_align of BITS_PER_UNIT.
24655 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24657         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24658         superfluous parentheses and trailing spaces.
24660 2018-02-28  Richard Biener  <rguenther@suse.de>
24662         PR tree-optimization/84584
24663         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24664         SCoPs with fake exit edge.
24666 2018-02-28  Martin Liska  <mliska@suse.cz>
24668         PR testsuite/84597
24669         * timevar.c (timer::print): Fix format to properly print 100%
24670         values.
24672 2018-02-28  Richard Biener  <rguenther@suse.de>
24674         PR middle-end/84607
24675         * genmatch.c (capture_info::walk_match): Do not mark
24676         captured expressions without operands as expr_p given
24677         they act more like predicates and should be subject to
24678         "lost tail" side-effect preserving.
24680 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24682         PR rtl-optimization/81611
24683         * auto-inc-dec.c (attempt_change): Move dead note from
24684         mem_insn if it's the next use of regno
24685         (find_address): Take address use of reg holding
24686         non-incremented value.  Add parm to limit search to the named
24687         reg only.
24688         (merge_in_block): Attempt to use a mem insn that is the next
24689         use of the original regno.
24691 2018-02-27  Martin Sebor  <msebor@redhat.com>
24693         PR c++/83871
24694         * doc/invoke.texi (-Wmissing-attributes): New option.
24695         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24697 2018-02-27  Martin Sebor  <msebor@redhat.com>
24699         PR translation/84207
24700         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24701         n argument to unsigned HOST_WIDE_INT.
24702         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24703         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24704         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24705         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24707 2018-02-27  Richard Biener  <rguenther@suse.de>
24709         PR tree-optimization/84512
24710         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24711         Do not use the estimate returned from record_stmt_cost for
24712         the scalar iteration cost but sum properly using add_stmt_cost.
24714 2018-02-27  Richard Biener  <rguenther@suse.de>
24716         PR tree-optimization/84466
24717         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24718         Adjust last change to less strictly validate use operands.
24720 2018-02-27  Martin Liska  <mliska@suse.cz>
24722         PR gcov-profile/84548
24723         * gcov.c (process_file): Allow partial overlap and consider it
24724         also as group functions.
24725         (output_lines): Properly calculate range of lines for a group.
24727 2018-02-27  Martin Liska  <mliska@suse.cz>
24729         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24730         'ggc' suffixes.  Change first column width.
24731         (timer::print): Fix formatting of the column.
24733 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24735         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24736         preserve inline entry blocks for the sake of debug inline
24737         entry point markers alone.
24738         (remove_unused_locals): Suggest in comments a better place to
24739         force the preservation of inline entry blocks that are
24740         otherwise unused, but do not preserve them.
24742 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24744         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24746 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24748         PR target/84039
24749         * config/i386/constraints.md (Bs): Replace
24750         ix86_indirect_branch_register with
24751         TARGET_INDIRECT_BRANCH_REGISTER.
24752         (Bw): Likewise.
24753         * config/i386/i386.md (indirect_jump): Likewise.
24754         (tablejump): Likewise.
24755         (*sibcall_memory): Likewise.
24756         (*sibcall_value_memory): Likewise.
24757         Peepholes of indirect call and jump via memory: Likewise.
24758         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24759         (*sibcall_value_GOT_32): Likewise.
24760         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24761         (GOT_memory_operand): Likewise.
24762         (call_insn_operand): Likewise.
24763         (sibcall_insn_operand): Likewise.
24764         (GOT32_symbol_operand): Likewise.
24765         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24767 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24769         PR rtl-optimization/83496
24770         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24771         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24772         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24773         redundant insn, if any.
24774         (relax_delay_slots): Likewise.
24775         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24777 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24779         PR tree-optimization/83965
24780         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24781         that grouped statements are part of a reduction chain.  Return
24782         true if the statement is not marked as a reduction itself but
24783         is part of a group.
24784         (vect_recog_dot_prod_pattern): Don't check whether the statement
24785         is part of a group here.
24786         (vect_recog_sad_pattern): Likewise.
24787         (vect_recog_widen_sum_pattern): Likewise.
24789 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24791         PR debug/84545
24792         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24794 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24796         PR target/84530
24797         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24798         the bool argument.
24799         (ix86_output_indirect_function_return): New prototype.
24800         (ix86_split_simple_return_pop_internal): Likewise.
24801         * config/i386/i386.c (indirect_return_via_cx): New.
24802         (indirect_return_via_cx_bnd): Likewise.
24803         (indirect_thunk_name): Handle return va CX_REG.
24804         (output_indirect_thunk_function): Create alias for
24805         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24806         (ix86_output_indirect_jmp): Remove the bool argument.
24807         (ix86_output_indirect_function_return): New function.
24808         (ix86_split_simple_return_pop_internal): Likewise.
24809         * config/i386/i386.md (*indirect_jump): Don't pass false
24810         to ix86_output_indirect_jmp.
24811         (*tablejump_1): Likewise.
24812         (simple_return_pop_internal): Change it to define_insn_and_split.
24813         Call ix86_split_simple_return_pop_internal to split it for
24814         -mfunction-return=.
24815         (simple_return_indirect_internal): Call
24816         ix86_output_indirect_function_return instead of
24817         ix86_output_indirect_jmp.
24819 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24821         PR bootstrap/84405
24822         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24823         memset and value initialization afterwards.
24825 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24827         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24829 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24831         PR target/84521
24832         * common/config/aarch64/aarch64-common.c
24833         (aarch_option_optimization_table[]): Switch
24834         off fomit-frame-pointer
24836 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24837             Chung-Ju Wu  <jasonwucj@gmail.com>
24839         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24840         volatile memory.
24841         (store_multiple): Ditto.
24843 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24845         * config.gcc: Add --with-cpu support for nds32 target.
24846         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24847         * config/nds32/nds32.opt: Add -mcpu= option.
24849 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24851         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24852         isel=yes): Warn for these deprecated options.
24854 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24856         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24857         ISA_2_5_MASKS_EMBEDDED.
24859 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24861         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24862         p->max as pointers rather than using iterative_hash_expr.
24864 2018-02-23  Carl Love  <cel@us.ibm.com>
24866         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24867         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24868         BU_P8V_OVERLOAD_2.
24869         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24870         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24871         P8V_BUILTIN_VEC_VUNSIGNED2.
24873 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
24875         PR target/81572
24876         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24877         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24878         LRA_UNKNOWN_ALT.
24879         * lra-constraints.c (curr_insn_transform): Set up
24880         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
24881         LRA_UNKNOWN_ALT.
24882         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24883         * lra-eliminations.c (spill_pseudos): Ditto.
24884         (process_insn_for_elimination): Ditto.
24885         * lra-lives.c (reg_early_clobber_p): Use the new macros.
24886         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24887         LRA_NON_CLOBBERED_ALT.
24889 2018-02-22  Martin Sebor  <msebor@redhat.com>
24891         PR tree-optimization/84480
24892         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24893         to maybe_diag_stxncpy_trunc.  Call it.
24894         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24895         from gimple_fold_builtin_strcpy.  Print inlining stack.
24896         (handle_builtin_stxncpy): Print inlining stack.
24897         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24899 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
24901         PR target/84176
24902         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24903         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24904         and -fcheck-pointer-bounds are used together.
24905         (indirect_thunk_prefix): New enum.
24906         (indirect_thunk_need_prefix): New function.
24907         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
24908         "_nt" instead of "_bnd" for NOTRACK prefix.
24909         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24910         (output_indirect_thunk_function): Likewise.
24911         (): Likewise.
24912         (ix86_code_end): Update output_indirect_thunk_function calls.
24913         (ix86_output_indirect_branch_via_reg): Replace
24914         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24915         (ix86_output_indirect_branch_via_push): Likewise.
24916         (ix86_output_function_return): Likewise.
24917         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24918         incompatible with -fcf-protection=branch and
24919         -fcheck-pointer-bounds.
24921 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24923         PR target/83335
24924         * config/aarch64/aarch64.c (aarch64_print_address_internal):
24925         Change gcc_assert call to output_operand_lossage.
24927 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
24929         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24931 2018-02-22  DJ Delorie  <dj@redhat.com>
24932             Sebastian Perta  <sebastian.perta@renesas.com>
24933             Oleg Endo  <olegendo@gcc.gnu.org>
24935         * config/rx/rx.c (rx_rtx_costs): New function.
24936         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24938 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24940         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24942 2018-02-22  Martin Liska  <mliska@suse.cz>
24944         PR driver/83193
24945         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24946         Add "native" as a possible value.
24948 2018-02-22  Martin Liska  <mliska@suse.cz>
24950         PR driver/83193
24951         * config/i386/i386.c (ix86_option_override_internal):
24952         Add "native" as a possible value for -march and -mtune.
24954 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
24956         PR target/84502
24957         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24958         to all type variants.
24960         PR tree-optimization/84503
24961         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24962         width as info->bitpos + info->bitsize - start.
24963         (merged_store_group::merge_overlapping): Simplify width computation.
24964         (check_no_overlap): New function.
24965         (imm_store_chain_info::try_coalesce_bswap): Compute expected
24966         start + width and last_order of the group, fail if check_no_overlap
24967         fails.
24968         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24969         to group if check_no_overlap fails.
24971 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24973         * config/rs6000/altivec.md: Delete contraint arguments to
24974         define_expand, define_split, and define_peephole2, and in
24975         define_insn_and_split if always unused.
24976         * config/rs6000/darwin.md: Ditto.
24977         * config/rs6000/dfp.md: Ditto.
24978         * config/rs6000/rs6000.md: Ditto.
24979         * config/rs6000/sync.md: Ditto.
24980         * config/rs6000/vector.md: Ditto.
24981         * config/rs6000/vsx.md: Ditto.
24983 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
24985         * config/rs6000/altivec.md: Write output control strings as braced
24986         blocks instead of double-quoted strings.
24987         * config/rs6000/darwin.md: Ditto.
24988         * config/rs6000/rs6000.md: Ditto.
24989         * config/rs6000/vector.md: Ditto.
24990         * config/rs6000/vsx.md: Ditto.
24992 2018-02-21  Jason Merrill  <jason@redhat.com>
24994         PR c++/84314 - ICE with templates and fastcall attribute.
24995         * attribs.c (build_type_attribute_qual_variant): Remove assert.
24997 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
24999         * ipa-cp.c (determine_versionability): Fix comment typos.
25001 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25003         PR c/84229
25004         * ipa-cp.c (determine_versionability): Do not version functions caling
25005         va_arg_pack.
25007 2018-02-21  Martin Liska  <mliska@suse.cz>
25009         PR driver/83193
25010         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25011         Add "native" as a possible value.
25012         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25013         the macro when native cpu detection is available.
25015 2018-02-21  Martin Liska  <mliska@suse.cz>
25017         PR driver/83193
25018         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25019         Add "native" as a possible value.
25020         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25021         when native cpu detection is available.
25023 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25024             Martin Sebor  <msebor@redhat.com>
25026         PR tree-optimization/84478
25027         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25028         false.
25029         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25030         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25031         support which is conservatively correct, for 2 only stay conservative
25032         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25033         argument to the 2 argument get_range_strlen, adjust 6 arg
25034         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25035         false.
25036         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25037         (gimple_fold_builtin_strlen): Pass true as last argument to
25038         get_range_strlen.
25040 2018-02-20  Martin Sebor  <msebor@redhat.com>
25042         PR middle-end/84095
25043         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25044         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25045         (builtin_memref::builtin_memref): Factor out parts into
25046         set_base_and_offset and call it.
25048 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25050         PR middle-end/84406
25051         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25052         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25053         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25054         search at the associated MODE_INT.
25056 2018-02-20  Jeff Law  <law@redhat.com>
25058         PR middle-end/82123
25059         PR tree-optimization/81592
25060         PR middle-end/79257
25061         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25062         for range data rather than using global data.
25063         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25064         range data rather than using global data.
25065         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25066         pass it to children as needed.
25067         (struct directive::fmtresult): Similarly.
25068         (struct directive::set_width): Similarly.
25069         (struct directive::set_precision): Similarly.
25070         (format_integer, format_directive, parse_directive): Similarly.
25071         (format_none): Accept unnamed vr_values parameter.
25072         (format_percent, format_floating, format_character): Similarly.
25073         (format_string, format_plain): Similarly.
25074         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25075         the EVRP range analyzer for range data rather than using global data.
25076         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25077         gimple-ssa-evrp-analyze.h
25078         (class sprintf_dom_walker): Add after_dom_children member function.
25079         Add evrp_range_analyzer member.
25080         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25081         range analyzer as needed.
25082         (sprintf_dom_walker::after_dom_children): New member function.
25083         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25084         if not optimizing.
25085         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25086         (evrp_range_analyzer::pop_to_marker): Likewise.
25088 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25090         PR tree-optimization/84419
25091         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25092         with the required type if its current type is compatible but
25093         different.
25095 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25097         PR middle-end/82004
25098         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25099         after vectorization.
25101 2018-02-20  Martin Liska  <mliska@suse.cz>
25103         PR driver/83193
25104         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25105         possible values if we don't have a hint.
25107 2018-02-20  Martin Liska  <mliska@suse.cz>
25109         PR c/84310
25110         PR target/79747
25111         * final.c (shorten_branches): Build align_tab array with one
25112         more element.
25113         * opts.c (finish_options): Add alignment option limit check.
25114         (MAX_CODE_ALIGN): Likewise.
25115         (MAX_CODE_ALIGN_VALUE): Likewise.
25116         * doc/invoke.texi: Document maximum allowed option value for
25117         all -falign-* options.
25119 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25121         PR target/84146
25122         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25123         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25124         * var-tracking.c (emit_note_insn_var_location): Remove all references
25125         to NOTE_INSN_CALL_ARG_LOCATION.
25126         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25127         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25128         Use copy_rtx_if_shared.
25129         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25130         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25131         (dwarf2out_var_location): Remove handling of
25132         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25133         on call_insn.
25134         * final.c (final_scan_insn): Remove all references to
25135         NOTE_INSN_CALL_ARG_LOCATION.
25136         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25137         before dumping final insns.
25138         * except.c (emit_note_eh_region_end): Remove all references to
25139         NOTE_INSN_CALL_ARG_LOCATION.
25140         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25141         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25142         * config/arc/arc.c (hwloop_optimize): Likewise.
25143         * config/arm/arm.c (create_fix_barrier): Likewise.
25144         * config/s390/s390.c (s390_chunkify_start): Likewise.
25145         * config/sh/sh.c (find_barrier): Likewise.
25146         * config/i386/i386.c (rest_of_insert_endbranch,
25147         ix86_seh_fixup_eh_fallthru): Likewise.
25148         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25149         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25150         * config/frv/frv.c (frv_function_prologue): Likewise.
25151         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25152         reg note.
25153         (note_outside_basic_block_p): Remove all references to
25154         NOTE_INSN_CALL_ARG_LOCATION.
25155         * gengtype.c (adjust_field_rtx_def): Likewise.
25156         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25157         Likewise.
25158         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25159         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25161         PR c++/84444
25162         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25163         is ADDR_EXPR.
25165         PR tree-optimization/84452
25166         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25167         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25168         is NULL.
25170 2018-02-19  Martin Liska  <mliska@suse.cz>
25172         PR sanitizer/82183
25173         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25175 2018-02-19  Martin Liska  <mliska@suse.cz>
25176             Richard Sandiford  <richard.sandiford@linaro.org>
25178         PR tree-optimization/82491
25179         * gimple-fold.c (get_base_constructor): Make earlier bail out
25180         to prevent ubsan.
25182 2018-02-19  Carl Love  <cel@us.ibm.com>
25184         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25185         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25186         BU_P8V_OVERLOAD_1.
25187         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25188         P8V_BUILTIN_VEC_NEG.
25190 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25192         * config/rl78/rl78.md (movdf): New define expand.
25194 2018-02-19  Martin Liska  <mliska@suse.cz>
25196         PR other/80589
25197         * doc/invoke.texi: Fix typo.
25198         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25200 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25202         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25203         handle rs6000_single_float and rs6000_double_float specially for
25204         e500 family CPUs.
25206 2018-02-16  Jeff Law  <law@redhat.com>
25208         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25209         (pop_regs): Use it.
25211 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25213         PR ipa/84425
25214         * ipa-inline.c (inline_small_functions): Fix a typo.
25216 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25218         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25220 2018-02-16  Carl Love  <cel@us.ibm.com>
25222         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25223         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25224         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25225         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25226         expansion to P8V_BUILTIN_VEC_FLOAT2.
25228 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25230         PR rtl-optimization/70023
25231         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25232         src_regno into account.
25234 2018-02-16  Carl Love  <cel@us.ibm.com>
25236         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25237         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25238         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25239         * config/rs6000/rs6000.c: Remove case statements for
25240         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25241         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25242         and P9V_BUILTIN_VEC_VINSERT4B.
25243         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25244         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25245         * config/rs6000/vsx.md:
25246         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25247         vec_insert4b.
25249 2018-02-16  Carl Love  <cel@us.ibm.com>
25251         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25252         vec_insert4b.
25253         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25254         definitions.
25255         * config/rs6000/rs6000-c.c: Add the definitions for
25256         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25257         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25258         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25259         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25260         definition for insert4b and define insn *insert3b_internal.
25261         * doc/extend.texi: Add documentation for vec_extract4b.
25263 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25265         * doc/extend.texi (Backwards Compatibility): Mention friend
25266         injection.  Note for-scope is deprecated.
25267         * doc/invoke.texi (-ffriend-injection): Deprecate.
25269 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25271         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25272         that moved to I2, also allow destinations that are a paradoxical
25273         subreg (instead of a normal reg).
25275 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25277         PR target/83831
25278         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25279         to QImode.
25281 2018-02-16  Richard Biener  <rguenther@suse.de>
25283         PR tree-optimization/84037
25284         PR tree-optimization/84016
25285         PR target/82862
25286         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25287         Adjust vec_construct for the fact we need additional higher latency
25288         128bit inserts for AVX256 and AVX512 vector builds.
25289         (ix86_add_stmt_cost): Scale vector construction cost for
25290         elementwise loads.
25292 2018-02-16  Richard Biener  <rguenther@suse.de>
25294         PR tree-optimization/84417
25295         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25296         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25297         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25299 2018-02-16  Martin Liska  <mliska@suse.cz>
25301         PR sanitizer/84307
25302         * internal-fn.def (ASAN_CHECK): Set proper flags.
25303         (ASAN_MARK): Likewise.
25305 2018-02-16  Julia Koval  <julia.koval@intel.com>
25307         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25308         from PTA_CANNONLAKE.
25310 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25312         PR target/84272
25313         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25314         Use ++iter rather than iter++ for std::list iterators.
25315         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25316         defer deleting them until all nodes in the forest are processed.  Do
25317         free even leaf nodes.  Change to_process into auto_vec.
25319         PR bootstrap/84405
25320         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25321         * vec.h (vec_default_construct): Use memset instead of placement new
25322         if BROKEN_VALUE_INITIALIZATION is defined.
25323         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25324         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25325         is defined.
25327         PR rtl-optimization/83723
25328         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25329         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25330         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25331         recursive calls.
25332         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25333         callers.
25334         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25336 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25338         PR rtl-optimization/81443
25339         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25340         from inner REGs to paradoxical SUBREGs.
25342 2018-02-16  Richard Biener  <rguenther@suse.de>
25344         PR tree-optimization/84399
25345         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25346         For operands we can analyze at their definition make sure we can
25347         analyze them at each use as well.
25349 2018-02-16  Richard Biener  <rguenther@suse.de>
25351         PR tree-optimization/84190
25352         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25353         volatile accesses if the decl isn't volatile.
25355 2018-02-15  Jason Merrill  <jason@redhat.com>
25357         PR c++/84314 - ICE with templates and fastcall attribute.
25358         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25359         TYPE_CANONICAL on an existing type.
25361 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25363         PR tree-optimization/84383
25364         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25365         dstoff nor call operand_equal_p if dstbase is NULL.
25367         PR tree-optimization/84334
25368         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25369         also a CONSTANT_CLASS_P, punt.
25371 2018-02-14  Jim Wilson  <jimw@sifive.com>
25373         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25374         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25375         to where locals are set.  Add TARGET_RVC support.
25376         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25378 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25380         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25382 2018-02-14  Martin Sebor  <msebor@redhat.com>
25384         PR tree-optimization/83698
25385         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25386         arrays constrain the offset range to their bounds.
25387         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25388         (builtin_access::overlap): Avoid setting the size of overlap if it's
25389         already been set.
25390         (maybe_diag_overlap): Also consider arrays when deciding what values
25391         of offsets to include in diagnostics.
25393 2018-02-14  Martin Sebor  <msebor@redhat.com>
25395         PR c/84108
25396         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25397         that correspond to the kind of a declaration.
25399 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25401         PR target/83984
25402         * config/pa/pa.md: Load address of PIC label using the linkage table
25403         if the label is nonlocal.
25405 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25407         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25408         warning message if user requests -maltivec=be.
25409         * doc/invoke.texi: Document deprecation of -maltivec=be.
25411 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25413         PR target/84220
25414         * config/rs6000/rs6000-c.c: Update definitions for
25415         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25416         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25418 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25420         PR target/84239
25421         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25422         add _get_ssp intrinsics. Remove argument from
25423         __builtin_ia32_rdssp[d|q].
25424         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25425         * config/i386/i386-builtin.def: Remove argument from
25426         __builtin_ia32_rdssp[d|q].
25427         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25428         ix86_expand_special_args_builtin for _rdssp[d|q].
25429         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25430         Clear register before usage.
25431         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25432         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25434 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25436         PR tree-optimization/84357
25437         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25438         operand 1 of an ARRAY_REF too.
25440 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25442         PR target/83831
25443         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25444         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25445         declarations.
25446         (set_of_reg): New struct.
25447         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25448         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25449         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25450         functions.
25451         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25452         Split into bitclr, bitset, bitinvert patterns if appropriate.
25453         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25454         use rx_fuse_in_memory_bitop.
25455         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25456         to named insn, correct maximum insn length.
25458 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25460         PR target/79242
25461         * machmode.def: Define a complex mode for PARTIAL_INT.
25462         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25463         MODE_PARTIAL_INT.
25464         * doc/rtl.texi: Document CSPImode.
25465         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25466         handling.
25467         (msp430_hard_regno_nregs_with_padding): Likewise.
25469 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25471         PR target/84279
25472         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25474 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25476         PR rtl-optimization/84169
25477         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25478         we generated a parallel as new i3 and we split that to new i2 and i3
25479         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25480         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25481         those to i2, not i1.  Partially rewrite this scan code.
25483 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25485         PR c/82210
25486         * stor-layout.c (place_field): For variable length fields, adjust
25487         offset_align afterwards not just based on the field's alignment,
25488         but also on the size.
25490         PR middle-end/84309
25491         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25492         of exps and logs in the use_exp2 case.
25494 2018-02-13  Jeff Law  <law@redhat.com>
25496         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25497         entry for "vector".
25499         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25500         ARGS as unused.
25502 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25504         PR debug/84342
25505         PR debug/84319
25506         * common.opt (gas-loc-support, gas-locview-support): New.
25507         (ginline-points, ginternal-reset-location-views): New.
25508         * doc/invoke.texi: Document them.  Use @itemx where intended.
25509         (gvariable-location-views): Adjust.
25510         * target.def (reset_location_view): New.
25511         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25512         (TARGET_RESET_LOCATION_VIEW): New.
25513         * doc/tm.texi: Rebuilt.
25514         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25515         (dwarf2out_default_as_locview_support): New.
25516         (output_asm_line_debug_info): Use option variables.
25517         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25518         (output_loc_list): Likewise.
25519         (add_high_low_attributes): Check option variables.
25520         Don't output entry view attribute in strict mode.
25521         (gen_inlined_subroutine_die): Check option variables.
25522         (dwarf2out_inline_entry): Likewise.
25523         (init_sections_and_labels): Likewise.
25524         (dwarf2out_early_finish): Likewise.
25525         (maybe_reset_location_view): New, from...
25526         (dwarf2out_var_location): ... here.  Call it.
25527         * debug.h (dwarf2out_default_as_loc_support): Declare.
25528         (dwarf2out_default_as_locview_support): Declare.
25529         * hooks.c (hook_int_rtx_insn_0): New.
25530         * hooks.h (hook_int_rtx_insn_0): Declare.
25531         * toplev.c (process_options): Take -gas-loc-support and
25532         -gas-locview-support from dwarf2out.  Enable
25533         -gvariable-location-views by default only with locview
25534         assembler support.  Enable -ginternal-reset-location-views by
25535         default only if the target defines the corresponding hook.
25536         Enable -ginline-points by default if location views are
25537         enabled; force it disabled if statement frontiers are
25538         disabled.
25539         * tree-inline.c (expand_call_inline): Check option variables.
25540         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25542 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25544         PR tree-optimization/84321
25545         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25546         handling.  Also check whether the anti-range contains any values
25547         that satisfy the mask; switch to a VR_RANGE if not.
25549 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25551         PR sanitizer/84340
25552         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25554 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25556         PR c++/83990
25557         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25558         of call statements, also set location of a load to a temporary.
25560 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25562         * config/rl78/rl78.c (add_vector_labels): New function.
25563         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25564         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25565         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25566         which checks that no arguments are passed.
25567         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25568         * doc/extend.texi: Documentation for the new attribute.
25570 2018-02-13  Andreas Schwab  <schwab@suse.de>
25572         * config/riscv/linux.h (CPP_SPEC): Define.
25574 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25576         PR target/84335
25577         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25578         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25579         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25580         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25581         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25582         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25583         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25584         temporarily for AES and PCLMUL builtins.
25586         PR tree-optimization/84339
25587         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25588         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25589         Formatting fixes.
25591         PR middle-end/84309
25592         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25593         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25594         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25595         inline function.
25596         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25597         inline function.
25598         * omp-simd-clone.h: New file.
25599         * omp-simd-clone.c: Include omp-simd-clone.h.
25600         (expand_simd_clones): No longer static.
25601         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25602         cgraph.h and omp-simd-clone.h.
25603         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25604         (vect_recog_widen_shift_pattern): Formatting fix.
25605         (vect_pattern_recog_1): Don't check optab for calls.
25607         PR target/84336
25608         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25609         operands[2] into a REG before using gen_lowpart on it.
25611 2018-02-12  Jeff Law  <law@redhat.com>
25613         PR target/83760
25614         * config/sh/sh.c (find_barrier): Consider a sibling call
25615         a barrier as well.
25617         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25618         successfully back substituting a reg.
25620 2018-02-12  Richard Biener  <rguenther@suse.de>
25622         PR tree-optimization/84037
25623         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25624         parameter, move visited init to caller.
25625         (vect_slp_analyze_operations): Separate cost from validity
25626         check, initialize visited once for all instances.
25627         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25628         for all instances.
25629         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25630         out an assert.
25631         (vect_model_promotion_demotion_cost): Likewise.
25632         (vectorizable_bswap): Guard cost modeling with !slp_node
25633         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25634         SLP stmts.
25635         (vectorizable_call): Likewise.
25636         (vectorizable_conversion): Likewise.
25637         (vectorizable_assignment): Likewise.
25638         (vectorizable_shift): Likewise.
25639         (vectorizable_operation): Likewise.
25640         (vectorizable_store): Likewise.
25641         (vectorizable_load): Likewise.
25642         (vectorizable_condition): Likewise.
25643         (vectorizable_comparison): Likewise.
25645 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25647         PR sanitizer/84307
25648         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25649         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25650         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25652 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25654         PR middle-end/83665
25655         * params.def (inline-min-speedup): Increase from 8 to 15.
25656         (max-inline-insns-auto): Decrease from 40 to 30.
25657         * ipa-split.c (consider_split): Add some buffer for function to
25658         be considered inlining candidate.
25659         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25660         default values.
25662 2018-02-12  Richard Biener  <rguenther@suse.de>
25664         PR tree-optimization/84037
25665         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25666         matched stmts if we cannot swap the non-matched ones.
25668 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25670         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25671         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25672         _mm_maskz_scalef_round_ss): New intrinsics.
25673         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25674         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25675         __builtin_ia32_scalefss_round): Remove.
25676         (__builtin_ia32_scalefsd_mask_round,
25677         __builtin_ia32_scalefss_mask_round): New intrinsics.
25678         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25679         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25680         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25681         "<round_constraint>")): Changed to ...
25682         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25683         "<round_scalar_constraint>")): ... this.
25684         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25685         %0, %1, %2<round_op3>}"): Changed to ...
25686         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25687         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25688         %2<round_scalar_mask_op3>}"): ... this.
25689         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25691 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25693         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25694         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25695         (_mm_maskz_sqrt_round_ss): New intrinsics.
25696         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25697         (__builtin_ia32_sqrtsd_mask_round)
25698         (__builtin_ia32_sqrtss_mask_round): New builtins.
25699         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25700         (__builtin_ia32_sqrtss_round): Remove.
25701         (__builtin_ia32_sqrtsd_mask_round)
25702         (__builtin_ia32_sqrtss_mask_round): New builtins.
25703         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25704         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25705         ((match_operand:VF_128 1 "vector_operand"
25706         "xBm,<round_constraint>")): Changed to ...
25707         ((match_operand:VF_128 1 "vector_operand"
25708         "xBm,<round_scalar_constraint>")): ... this.
25709         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25710         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25711         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25712         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25713         %<iptr>1<round_scalar_mask_op3>}): ... this.
25714         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25715         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25717 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25719         PR target/84266
25720         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25721         Cast vec_cmpeq result to correct type.
25722         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25723         Cast vec_cmpgt result to correct type.
25725 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25727         * final.c (final_scan_insn_1): Renamed from...
25728         (final_scan_insn): ... this.  New wrapper, to recover
25729         seen from the outermost call in recursive ones.
25730         * config/sparc/sparc.c (output_return): Drop seen from call.
25731         (output_sibcall): Likewise.
25732         * config/visium/visium.c (output_branch): Likewise.
25734 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25736         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25737         function label.
25739 2018-02-10  Alan Modra  <amodra@gmail.com>
25741         PR target/84300
25742         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25743         Specify LR as an input.
25745 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25747         PR sanitizer/83987
25748         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25749         remove_member_access_dummy_vars): New functions.
25750         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25751         lower_omp_1, execute_lower_omp): Use them.
25753         PR rtl-optimization/84308
25754         * shrink-wrap.c (spread_components): Release todo vector.
25756 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25758         PR rtl-optimization/57193
25759         * ira-color.c (struct allocno_color_data): Add member
25760         conflict_allocno_hard_prefs.
25761         (update_conflict_allocno_hard_prefs): New.
25762         (bucket_allocno_compare_func): Add a preference based on
25763         conflict_allocno_hard_prefs.
25764         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25765         (color_allocnos): Remove a dead code.  Initiate
25766         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25768 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25770         PR target/84226
25771         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25772         constraint from =wa to wa.  Avoid a subreg on the output operand,
25773         instead use a pseudo and subreg it in a move.
25774         (p9_xxbrd_<mode>): Changed to ...
25775         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25776         (p9_xxbrd_v2df): New expander.
25777         (p9_xxbrw_<mode>): Changed to ...
25778         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25779         (p9_xxbrw_v4sf): New expander.
25781 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25783         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25785 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25787         PR target/83926
25788         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25789         multiply in 32-bit mode.
25790         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25791         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25792         mode.
25794 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25796         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25797         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25798         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25799         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25801 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25803         PR lto/84213
25804         * dwarf2out.c (is_trivial_indirect_ref): New function.
25805         (dwarf2out_late_global_decl): Do not generate a location
25806         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25807         and that are not defined in the current unit.
25809 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25811         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25812         instead of a libcall for UNORDERED.
25814 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25816         PR target/82641
25817         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25818         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25820 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25822         PR target/PR84295
25823         * config/s390/s390.c (s390_set_current_function): Invoke
25824         s390_indirect_branch_settings also if fndecl didn't change.
25826 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25828         * config/rs6000/rs6000.md (blockage): Set length to zero.
25830 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25832         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25834 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25836         PR sanitizer/84285
25837         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25838         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25839         -static-lib*san.
25841         PR debug/84252
25842         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25843         PARALLEL incoming that failed vt_get_decl_and_offset check.
25845         PR middle-end/84237
25846         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25847         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25848         TREE_READONLY bit.
25849         (get_variable_section): For decls in named .bss* sections pass true as
25850         second argument to bss_initializer_p.
25852 2018-02-09  Marek Polacek  <polacek@redhat.com>
25853             Jakub Jelinek  <jakub@redhat.com>
25855         PR c++/83659
25856         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25857         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25858         Sync some changes from cxx_fold_indirect_ref.
25860 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25862         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25863         markers.
25864         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25865         (BLOCK_INLINE_ENTRY_LABEL): New.
25866         (dwarf2out_var_location): Disregard inline entry markers.
25867         (inline_entry_data): New struct.
25868         (inline_entry_data_hasher): New hashtable type.
25869         (inline_entry_data_hasher::hash): New.
25870         (inline_entry_data_hasher::equal): New.
25871         (inline_entry_data_table): New variable.
25872         (add_high_low_attributes): Add DW_AT_entry_pc and
25873         DW_AT_GNU_entry_view attributes if a pending entry is found
25874         in inline_entry_data_table.  Add old entry_pc attribute only
25875         if debug nonbinding markers are disabled.
25876         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25877         markers are enabled.
25878         (block_within_block_p, dwarf2out_inline_entry): New.
25879         (dwarf2out_finish): Check that no entries remained in
25880         inline_entry_data_table.
25881         * final.c (reemit_insn_block_notes): Handle inline entry notes.
25882         (final_scan_insn, notice_source_line): Likewise.
25883         (rest_of_clean_state): Skip inline entry markers.
25884         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25885         markers.
25886         * gimple.c (gimple_build_debug_inline_entry): New.
25887         * gimple.h (enum gimple_debug_subcode): Add
25888         GIMPLE_DEBUG_INLINE_ENTRY.
25889         (gimple_build_debug_inline_entry): Declare.
25890         (gimple_debug_inline_entry_p): New.
25891         (gimple_debug_nonbind_marker_p): Adjust.
25892         * insn-notes.def (INLINE_ENTRY): New.
25893         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25894         inline entry marker notes.
25895         (print_insn): Likewise.
25896         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25897         (INSN_DEBUG_MARKER_KIND): Likewise.
25898         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25899         * tree-inline.c (expand_call_inline): Build and insert
25900         debug_inline_entry stmt.
25901         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25902         inline entry blocks early, if nonbind markers are enabled.
25903         (dump_scope_block): Dump fragment info.
25904         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25905         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25906         (gimple_build_debug_inline_entry): New.
25907         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25908         Enable/disable inline entry points too.
25909         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25910         (DEBUG_INSN): Describe inline entry markers.
25912         * common.opt (gvariable-location-views): New.
25913         (gvariable-location-views=incompat5): New.
25914         * config.in: Rebuilt.
25915         * configure: Rebuilt.
25916         * configure.ac: Test assembler for view support.
25917         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25918         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25919         * dwarf2out.c (var_loc_view): New typedef.
25920         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25921         (dwarf2out_locviews_in_attribute): New.
25922         (dwarf2out_locviews_in_loclist): New.
25923         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25924         (enum dw_line_info_opcode): Add LI_adv_address.
25925         (struct dw_line_info_table): Add view.
25926         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25927         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25928         (zero_view_p): New variable.
25929         (ZERO_VIEW_P): New macro.
25930         (output_asm_line_debug_info): New.
25931         (struct var_loc_node): Add view.
25932         (add_AT_view_list, AT_loc_list): New.
25933         (add_var_loc_to_decl): Add view param.  Test it against last.
25934         (new_loc_list): Add view params.  Record them.
25935         (AT_loc_list_ptr): Handle loc and view lists.
25936         (view_list_to_loc_list_val_node): New.
25937         (print_dw_val): Handle dw_val_class_view_list.
25938         (size_of_die): Likewise.
25939         (value_format): Likewise.
25940         (loc_list_has_views): New.
25941         (gen_llsym): Set vl_symbol too.
25942         (maybe_gen_llsym, skip_loc_list_entry): New.
25943         (dwarf2out_maybe_output_loclist_view_pair): New.
25944         (output_loc_list): Output view list or entries too.
25945         (output_view_list_offset): New.
25946         (output_die): Handle dw_val_class_view_list.
25947         (output_dwarf_version): New.
25948         (output_compilation_unit_header): Use it.
25949         (output_skeleton_debug_sections): Likewise.
25950         (output_rnglists, output_line_info): Likewise.
25951         (output_pubnames, output_aranges): Update version comments.
25952         (output_one_line_info_table): Output view numbers in asm comments.
25953         (dw_loc_list): Determine current endview, pass it to new_loc_list.
25954         Call maybe_gen_llsym.
25955         (loc_list_from_tree_1): Adjust.
25956         (add_AT_location_description): Create view list attribute if
25957         needed, check it's absent otherwise.
25958         (convert_cfa_to_fb_loc_list): Adjust.
25959         (maybe_emit_file): Call output_asm_line_debug_info for test.
25960         (dwarf2out_var_location): Reset views as needed.  Precompute
25961         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
25962         attribute.  Set view.
25963         (new_line_info_table): Reset next view.
25964         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25965         (dwarf2out_source_line): Likewise.  Output view resets and labels to
25966         the assembler, or select appropriate line info opcodes.
25967         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25968         (optimize_string_length): Catch it.  Adjust.
25969         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
25970         dw_val_class_view_list, and remove it if no longer needed.
25971         (hash_loc_list): Hash view numbers.
25972         (loc_list_hasher::equal): Compare them.
25973         (optimize_location_lists): Check whether a view list symbol is
25974         needed, and whether the locview attribute is present, and
25975         whether they match.  Remove the locview attribute if no longer
25976         needed.
25977         (index_location_lists): Call skip_loc_list_entry for test.
25978         (dwarf2out_finish): Call output_asm_line_debug_info for test.
25979         Use output_dwarf_version.
25980         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25981         (struct dw_val_node): Add val_view_list.
25982         * final.c (SEEN_NEXT_VIEW): New.
25983         (set_next_view_needed): New.
25984         (clear_next_view_needed): New.
25985         (maybe_output_next_view): New.
25986         (final_start_function): Rename to...
25987         (final_start_function_1): ... this.  Take pointer to FIRST,
25988         add SEEN parameter.  Emit param bindings in the initial view.
25989         (final_start_function): Reintroduce SEEN-less interface.
25990         (final): Rename to...
25991         (final_1): ... this.  Take SEEN parameter.  Output final pending
25992         next view at the end.
25993         (final): Reintroduce seen-less interface.
25994         (final_scan_insn): Output pending next view before switching
25995         sections or ending a block.  Mark the next view as needed when
25996         outputting variable locations.  Notify debug backend of section
25997         changes, and of location view changes.
25998         (rest_of_handle_final): Adjust.
25999         * toplev.c (process_options): Autodetect value for debug variable
26000         location views option.  Warn on incompat5 without -gdwarf-5.
26001         * doc/invoke.texi (gvariable-location-views): New.
26002         (gvariable-location-views=incompat5): New.
26003         (gno-variable-location-views): New.
26005 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26007         PR tree-optimization/84136
26008         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26009         that the result of find_edge is non-NULL.
26011 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26013         PR target/83008
26014         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26015         storing integer register in SImode.  Fix cost of 256 and 512
26016         byte aligned SSE register store.
26018 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26020         * config/i386/i386.c (ix86_multiplication_cost): Fix
26021         multiplication cost for TARGET_AVX512DQ.
26023 2018-02-08  Marek Polacek  <polacek@redhat.com>
26025         PR tree-optimization/84238
26026         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26027         get_range_strlen.
26029 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26031         PR tree-optimization/84265
26032         * tree-vect-stmts.c (vectorizable_store): Don't treat
26033         VMAT_CONTIGUOUS accesses as grouped.
26034         (vectorizable_load): Likewise.
26036 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26038         PR tree-optimization/81635
26039         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26040         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26041         (test_round_for_mask): New functions.
26042         (wide_int_cc_tests): Call test_round_for_mask.
26043         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26044         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26045         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26046         range returned by get_range_info.
26048 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26050         PR ipa/81360
26051         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26052         * symtab.c: Include builtins.h
26053         (symtab_node::output_to_lto_symbol_table_p): Move here
26054         from lto-streamer-out.c:output_symbol_p.
26055         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26056         (output_symbol_p): Move all logic to symtab.c
26057         (produce_symtab): Update.
26059 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26061         * config/s390/s390-opts.h (enum indirect_branch): Define.
26062         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26063         (s390_indirect_branch_via_thunk)
26064         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26065         (enum s390_indirect_branch_type): Define.
26066         * config/s390/s390.c (struct s390_frame_layout, struct
26067         machine_function): Remove.
26068         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26069         (indirect_branch_table_label_no, indirect_branch_table_name):
26070         Define variables.
26071         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26072         (enum s390_indirect_branch_option): Define.
26073         (s390_return_addr_from_memory): New function.
26074         (s390_handle_string_attribute): New function.
26075         (s390_attribute_table): Add new attribute handler.
26076         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26077         (s390_indirect_branch_via_thunk): New function.
26078         (s390_indirect_branch_via_inline_thunk): New function.
26079         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26080         sibling call optimization for non z10 compiles.
26081         (s390_emit_call): Force indirect branch target to be a single
26082         register.  Add r1 clobber for non-z10 compiles.
26083         (s390_emit_epilogue): Emit return jump via return_use expander.
26084         (s390_reorg): Handle JUMP_INSNs as execute targets.
26085         (s390_option_override_internal): Perform validity checks for the
26086         new command line options.
26087         (s390_indirect_branch_attrvalue): New function.
26088         (s390_indirect_branch_settings): New function.
26089         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26090         (s390_output_indirect_thunk_function):  New function.
26091         (s390_code_end): Implement target hook.
26092         (s390_case_values_threshold): Implement target hook.
26093         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26094         macros.
26095         * config/s390/s390.h (struct s390_frame_layout)
26096         (struct machine_function): Move here from s390.c.
26097         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26098         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26099         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26100         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26101         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26102         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26103         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26104         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26105         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26106         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26107         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26108         (mnemonic attribute): Add values which aren't recognized
26109         automatically.
26110         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26111         pattern for branch conversion.  Fix mnemonic attribute.
26112         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26113         indirect branch via thunk if requested.
26114         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26115         ("*indirect_jump"): Disable for branch conversion using out of
26116         line thunks.
26117         ("indirect_jump_via_thunk<mode>_z10")
26118         ("indirect_jump_via_thunk<mode>")
26119         ("indirect_jump_via_inlinethunk<mode>_z10")
26120         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26121         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26122         ("casesi_jump_via_inlinethunk<mode>_z10")
26123         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26124         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26125         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26126         ("*indirect2_jump"): Disable for branch conversion.
26127         ("casesi_jump"): Turn into expander and expand patterns for branch
26128         conversion.
26129         ("return_use"): New expander.
26130         ("*return"): Emit return via thunk and rename it to ...
26131         ("*return<mode>"): ... this one.
26132         * config/s390/s390.opt: Add new options and and enum for the
26133         option values.
26135 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26137         * lra-constraints.c (match_reload): Unconditionally use
26138         gen_lowpart_SUBREG, rather than selecting between that
26139         and equivalent gen_rtx_SUBREG code.
26141 2018-02-08  Richard Biener  <rguenther@suse.de>
26143         PR tree-optimization/84233
26144         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26145         changed flag instead of boguously re-using phi_inserted.
26147 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26149         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26150         static local variables.
26152 2018-02-08  Richard Biener  <rguenther@suse.de>
26154         PR tree-optimization/84278
26155         * tree-vect-stmts.c (vectorizable_store): When looking for
26156         smaller vector types to perform grouped strided loads/stores
26157         make sure the mode is supported by the target.
26158         (vectorizable_load): Likewise.
26160 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26162         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26163         Increase LDP/STP opportunities by adding adjacent callee-saves.
26165 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26167         PR rtl-optimization/84068
26168         PR rtl-optimization/83459
26169         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26171 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26173         PR tree-optimization/84224
26174         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26175         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26176         non-zero arguments.
26178 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26180         PR target/84113
26181         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26182         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26183         count, remove one USE.
26185 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26187         * doc/install.texi (Configuration): Document the
26188         --with-long-double-format={ibm,ieee} PowerPC configuration
26189         options.
26191         PR target/84154
26192         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26193         Convert from define_expand to be define_insn_and_split.  Rework
26194         float/double/_Float128 conversions to QI/HI/SImode to work with
26195         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26196         conversions to QI/HImode types did a store and then a load to
26197         truncate the value.  For conversions to VSX registers, don't split
26198         the insn, instead emit the code directly.  Use the code iterator
26199         any_fix to combine signed and unsigned conversions.
26200         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26201         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26202         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26203         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26204         (fix_<mode>di2_hw): Likewise.
26205         (fixuns_<mode>di2_hw): Likewise.
26206         (fix_<mode>si2_hw): Likewise.
26207         (fixuns_<mode>si2_hw): Likewise.
26208         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26209         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26210         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26211         fix<uns>_trunc<SFDF:mode>si2_p8.
26212         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26213         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26214         (fix<uns>_<mode>_mem): Likewise.
26215         (fctiw<u>z_<mode>_mem): Likewise.
26216         (fix<uns>_<mode>_mem): Likewise.
26217         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26218         the register allocator from doing a direct move to the GPRs to do
26219         a store, and instead use the ISA 3.0 store byte/half-word from
26220         vector register instruction.  For IEEE 128-bit floating point,
26221         also optimize stores of 32-bit ints.
26222         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26224 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26226         * genextract.c (push_pathstr_operand): New function to support
26227         [a-zA-Z].
26228         (walk_rtx): Call push_pathstr_operand.
26229         (print_path): Support [a-zA-Z].
26231 2018-02-07  Richard Biener  <rguenther@suse.de>
26233         PR tree-optimization/84037
26234         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26235         (cse_and_gimplify_to_preheader): Declare.
26236         (vect_get_place_in_interleaving_chain): Likewise.
26237         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26238         ivexpr_map.
26239         (_loop_vec_info::~_loop_vec_info): Delete it.
26240         (cse_and_gimplify_to_preheader): New function.
26241         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26242         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26243         (vectorizable_load): Likewise.  For grouped stores always base
26244         the IV on the first element.
26245         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26246         condition before gimplifying.
26248 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26250         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26251         *DIV_EXPR and *MOD_EXPR.
26253 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26255         PR target/84248
26256         * config/i386/i386.c (ix86_option_override_internal): Mask out
26257         the CF_SET bit when checking -fcf-protection.
26259 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26261         PR libgomp/84217
26262         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26263         enough.
26265 2018-02-07  Richard Biener  <rguenther@suse.de>
26267         PR tree-optimization/84204
26268         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26269         this place.
26271         PR tree-optimization/84205
26272         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26273         special-case isl_ast_op_zdiv_r.
26275         PR tree-optimization/84223
26276         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26277         Only add conditions from within the region.
26278         (gather_bbs::after_dom_children): Adjust.
26280 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26282         PR target/84209
26283         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26284         * config/avr/avr.md: Only post-reload split REG-REG moves if
26285         either register is GENERAL_REG_P.
26287 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26289         PR tree-optimization/84235
26290         * tree-ssa-scopedtables.c
26291         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26292         if the subtraction is performed in floating point type where NaNs are
26293         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26294         build 1.  Formatting fix.
26296 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26298         PR target/84146
26299         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26300         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26301         and skip it regardless of bb boundaries.  Use CALL_P macro,
26302         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26303         unnecessarily, formatting fix.
26305 2018-02-06  Michael Collison  <michael.collison@arm.com>
26307         * config/arm/thumb2.md:
26308         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26309         (*thumb_mov_notscc): Ditto.
26311 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26313         PR target/84154
26314         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26315         unsigned_fix, not "s".
26317 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26319         * configure.ac (gcc_fn_eh_frame_ro): New function.
26320         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26321         correct .eh_frame permissions.
26322         * configure: Regenerate.
26324 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26326         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26327         irrelevant options.
26329 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26331         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26332         Display warning message for -mno-speculate-indirect-jumps.
26334 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26336         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26337         Undocumented.
26338         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26340 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26342         PR tree-optimization/84225
26343         * tree-eh.c (find_trapping_overflow): Only call
26344         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26346 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26348         PR target/84145
26349         * config/i386/i386.c: Reimplement the check of possible options
26350         -mibt/-mshstk conbination. Change error messages.
26351         * doc/invoke.texi: Fix a typo: remove extra '='.
26353 2018-02-06  Marek Polacek  <polacek@redhat.com>
26355         PR tree-optimization/84228
26356         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26358 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26360         PR target/82641
26361         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26362         emitted arch directives.
26363         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26364         __ARM_FEATURE_COPROC before changing architectures.
26366 2018-02-06  Richard Biener  <rguenther@suse.de>
26368         * config/i386/i386.c (print_reg): Fix typo.
26369         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26371 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26373         * configure: Regenerate.
26375 2018-02-05  Martin Sebor  <msebor@redhat.com>
26377         PR tree-optimization/83369
26378         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26379         inlining context.
26381 2018-02-05  Martin Liska  <mliska@suse.cz>
26383         * doc/invoke.texi: Cherry-pick upstream r323995.
26385 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26387         * ira.c (ira_init_register_move_cost): Adjust comment.
26389 2018-02-05  Martin Liska  <mliska@suse.cz>
26391         PR gcov-profile/84137
26392         * doc/gcov.texi: Fix typo in documentation.
26394 2018-02-05  Martin Liska  <mliska@suse.cz>
26396         PR gcov-profile/83879
26397         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26398         using dlopen functionality.
26400 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26402         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26403         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26404         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26405         _mm_maskz_range_round_ss): New intrinsics.
26406         (__builtin_ia32_rangesd128_round)
26407         (__builtin_ia32_rangess128_round): Remove.
26408         (__builtin_ia32_rangesd128_mask_round,
26409         __builtin_ia32_rangess128_mask_round): New builtins.
26410         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26411         __builtin_ia32_rangess128_round): Remove.
26412         (__builtin_ia32_rangesd128_mask_round,
26413         __builtin_ia32_rangess128_mask_round): New builtins.
26414         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26415         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26416         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26417         "<round_saeonly_constraint>")): Changed to ...
26418         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26419         "<round_saeonly_scalar_constraint>")): ... this.
26420         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26421         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26422         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26423         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26424         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26426 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26428         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26429         options.
26430         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26431         Remove all values except native, 8540 and 8548.
26433 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26435         * config/i386/i386.c (ix86_output_function_return): Pass
26436         INVALID_REGNUM, instead of -1, as invalid register number to
26437         indirect_thunk_name and output_indirect_thunk.
26439 2018-02-02  Julia Koval  <julia.koval@intel.com>
26441         * config.gcc: Add -march=icelake.
26442         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26443         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26444         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26445         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26446         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26447         (processor_target_table): Add icelake.
26448         (ix86_option_override_internal): Handle new PTAs.
26449         (get_builtin_code_for_version): Handle icelake.
26450         (M_INTEL_COREI7_ICELAKE): New.
26451         (fold_builtin_cpu): Handle icelake.
26452         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26453         * doc/invoke.texi: Add -march=icelake.
26455 2018-02-02  Julia Koval  <julia.koval@intel.com>
26457         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26458         to wide_int_bitmask.
26459         * wide-int-bitmask.h: New.
26461 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26463         PR target/84066
26464         * config/i386/i386.md: Replace Pmode with word_mode in
26465         builtin_setjmp_setup and builtin_longjmp to support x32.
26467 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26469         PR target/56010
26470         PR target/83743
26471         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26472         #include "opts.h".
26473         (rs6000_supported_cpu_names): New static variable.
26474         (linux_cpu_translation_table): Likewise.
26475         (elf_platform) <cpu>: Define new static variable and use it.
26476         Translate kernel AT_PLATFORM name to canonical name if needed.
26477         Error if platform name is unknown.
26479 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26481         PR target/84089
26482         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26484 2018-02-01  Jeff Law  <law@redhat.com>
26486         PR target/84128
26487         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26488         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26489         the scratch if RELEASE_VIA_POP is false.
26490         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26491         If we have to save a temporary register, decrement SIZE appropriately.
26492         Pass new arguments to release_scratch_register_on_entry.
26493         (ix86_adjust_stack_and_probe): Likewise.
26494         (ix86_emit_probe_stack_range): Pass new arguments to
26495         release_scratch_register_on_entry.
26497 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26499         PR rtl-optimization/84157
26500         * combine.c (change_zero_ext): Use REG_P predicate in
26501         front of HARD_REGISTER_P predicate.
26503 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26505         * config/avr/avr.c (avr_option_override): Move disabling of
26506         -fdelete-null-pointer-checks to...
26507         * common/config/avr/avr-common.c (avr_option_optimization_table):
26508         ...here.
26510 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26512         PR tree-optimization/81635
26513         * tree-data-ref.c (split_constant_offset_1): For types that
26514         wrap on overflow, try to use range info to prove that wrapping
26515         cannot occur.
26517 2018-02-01  Renlin Li  <renlin.li@arm.com>
26519         PR target/83370
26520         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26521         TAILCALL_ADDR_REGS.
26522         (aarch64_register_move_cost): Likewise.
26523         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26524         TAILCALL_ADDR_REGS.
26525         (REG_CLASS_NAMES): Likewise.
26526         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26527         TAILCALL_ADDR_REGS. Remove IP registers.
26528         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26530 2018-02-01  Richard Biener  <rguenther@suse.de>
26532         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26533         for specifying RPO order and allow NULL for that.
26534         * domwalk.c (dom_walker::dom_walker): Likewise.
26535         (dom_walker::walk): Handle NULL RPO order.
26536         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26537         in RPO order.
26538         (rewrite_update_dom_walker): Likewise.
26539         (mark_def_dom_walker): Likewise.
26541 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26543         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26544         (aarch64_maybe_expand_sve_subreg_move): Declare.
26545         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26546         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26547         predicate.
26548         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26549         that are semantically a reverse operation.
26550         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26551         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26552         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26553         functions.
26554         (aarch64_can_change_mode_class): For big-endian, forbid changes
26555         between two SVE modes if they have different element sizes.
26557 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26559         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26560         the TImode handling for big-endian targets.
26562 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26564         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26565         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26566         not just bytes.
26567         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26568         Remove BSWAP handing for big-endian targets and use the form of
26569         LD1RQ appropariate for the mode.
26571 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26573         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26574         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26575         duplicated element.
26577 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26579         PR tearget/83845
26580         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26581         check for operands that need to go through aarch64_sve_reload_be.
26583 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26585         PR tree-optimization/81661
26586         PR tree-optimization/84117
26587         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26588         * tree-eh.c: Include gimplify.h.
26589         (find_trapping_overflow, replace_trapping_overflow,
26590         rewrite_to_non_trapping_overflow): New functions.
26591         * tree-vect-loop.c: Include tree-eh.h.
26592         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26593         * tree-data-ref.c: Include tree-eh.h.
26594         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26596 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26598         PR rtl-optimization/84123
26599         * combine.c (change_zero_ext): Check if hard register satisfies
26600         can_change_dest_mode before calling gen_lowpart_SUBREG.
26602 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26604         PR target/82444
26605         * ira.c (ira_init_register_move_cost): Remove assert.
26607 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26609         PR rtl-optimization/84071
26610         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26611         * doc/tm.texi: Regenerate.
26613 2018-01-31  Richard Biener  <rguenther@suse.de>
26615         PR tree-optimization/84132
26616         * tree-data-ref.c (analyze_miv_subscript): Properly
26617         check whether evolution_function_is_affine_multivariate_p
26618         before calling gcd_of_steps_may_divide_p.
26620 2018-01-31  Julia Koval  <julia.koval@intel.com>
26622         PR target/83618
26623         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26624         * config/i386/i386.md (rdpid_rex64) New.
26625         (rdpid): Make 32bit only.
26627 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26629         PR lto/84105
26630         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26631         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26633 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26635         Revert
26636         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26638         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26640 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26642         PR rtl-optimization/84071
26643         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26644         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26646 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26648         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26649         (arc_attribute_table): Add 'aux' attribute.
26650         (arc_in_small_data_p): Consider aux like variables.
26651         (arc_is_aux_reg_p): New function.
26652         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26653         (arc_get_aux_arg): New function.
26654         (prepare_move_operands): Handle aux-register access.
26655         (arc_handle_aux_attribute): New function.
26656         * doc/extend.texi (ARC Variable attributes): Add subsection.
26658 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26660         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26661         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26662         (arc_attribute_table): Add 'uncached' attribute.
26663         (arc_print_operand): Print '.di' flag for uncached memory
26664         accesses.
26665         (arc_in_small_data_p): Do not consider for small data the uncached
26666         types.
26667         (arc_is_uncached_mem_p): New function.
26668         * config/arc/predicates.md (compact_store_memory_operand): Check
26669         for uncached memory accesses.
26670         (nonvol_nonimm_operand): Likewise.
26671         * doc/extend.texi (ARC Type Attribute): New subsection.
26673 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26675         PR c/84100
26676         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26677         falign-loops=): Add Optimization flag.
26679 2018-01-30  Jeff Law  <law@redhat.com>
26681         PR target/84064
26682         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26683         INT_REGISTERS_SAVED.  Check it prior to calling
26684         get_scratch_register_on_entry.
26685         (ix86_adjust_stack_and_probe): Similarly.
26686         (ix86_emit_probe_stack_range): Similarly.
26687         (ix86_expand_prologue): Corresponding changes.
26689 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26691         PR target/40411
26692         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26693         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26695 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26697         PR target/84112
26698         * lra-constraints.c (curr_insn_transform): Process AND in the
26699         address.
26701 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26703         PR rtl-optimization/83986
26704         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26705         dependence against last_pending_memory_flush in addition to
26706         pending_jump_insns.
26708 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26710         PR tree-optimization/81611
26711         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26712         copies.
26714 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26716         PR target/83758
26717         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26718         a reg rtx.
26720 2018-01-30  Richard Biener  <rguenther@suse.de>
26721             Jakub Jelinek  <jakub@redhat.com>
26723         PR tree-optimization/84111
26724         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26725         inner loops added during recursion, as they don't have up-to-date
26726         SSA form.
26728 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26730         PR ipa/81360
26731         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26732         (can_inline_edge_by_limits_p): ... here.
26733         (can_early_inline_edge_p, check_callers,
26734         update_caller_keys, update_callee_keys, recursive_inlining,
26735         add_new_edges_to_heap, speculation_useful_p,
26736         inline_small_functions,
26737         inline_small_functions, flatten_function,
26738         inline_to_all_callers_1): Update.
26740 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26742         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26743         zeros correctly.
26745 2018-01-30  Richard Biener  <rguenther@suse.de>
26747         PR tree-optimization/83008
26748         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26749         invariant and constant vector uses in stmts when they need
26750         more than one stmt.
26752 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26754         PR bootstrap/84017
26755         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26756         * configure: Regenerate.
26758 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26760         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26761         pattern.
26762         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26763         Use gen_rtx_REG rather than gen_lowpart.
26765 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26767         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26768         rather than 0 when creating partial subregs.
26770 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26772         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26773         of usage.
26775 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26777         PR target/81550
26778         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26779         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26780         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26781         flags.  This restores the settings used before the 2017-07-24.
26782         Turning off pre increment/decrement/modify allows IVOPTS to
26783         optimize DF/SF loops where the index is an int.
26785 2018-01-29  Richard Biener  <rguenther@suse.de>
26786             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26788         PR bootstrap/80867
26789         * tree-vect-stmts.c (vectorizable_call): Don't call
26790         targetm.vectorize_builtin_md_vectorized_function if callee is
26791         NULL.
26793 2018-01-22  Carl Love  <cel@us.ibm.com>
26795         * doc/extend.tex: Fix typo in second arg in
26796         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26798 2018-01-29  Richard Biener  <rguenther@suse.de>
26800         PR tree-optimization/84086
26801         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26802         (flush_ssaname_freelist): When SSA names were released reset
26803         the SCEV hash table.
26805 2018-01-29  Richard Biener  <rguenther@suse.de>
26807         PR tree-optimization/84057
26808         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26809         removed paths when removing edges.
26811 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26813         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26814         -mfunction-return=@var{choice}.
26816 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26818         PR diagnostic/84034
26819         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26820         Handle CR like TAB.
26821         (layout::print_source_line): Likewise.
26822         (test_get_line_width_without_trailing_whitespace): Add test cases.
26824 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26826         PR middle-end/84040
26827         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26828         debug insns.
26830 2018-01-26  Jim Wilson  <jimw@sifive.com>
26832         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26834         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26835         specified.
26837 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26839         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26840         and CMP + SUB-immediate -> SUBS.
26842 2018-01-26  Martin Sebor  <msebor@redhat.com>
26844         PR tree-optimization/83896
26845         * tree-ssa-strlen.c (get_string_len): Rename...
26846         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26847         Avoid assuming length is constant.
26848         (handle_char_store): Use HOST_WIDE_INT for string length.
26850 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26852         PR target/81763
26853         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26854         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26856 2018-01-26  Richard Biener  <rguenther@suse.de>
26858         PR rtl-optimization/84003
26859         * dse.c (record_store): Only record redundant stores when
26860         the earlier store aliases at least all accesses the later one does.
26862 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26864         PR rtl-optimization/83985
26865         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26866         REG_CFA_RESTORE insns.
26867         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26868         REG_CFA_RESTORE insns here.
26870         PR c/83989
26871         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26872         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26874 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26876         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26877         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26878         (arc_init): Likewise.
26879         (arc_override_options): Likewise.
26880         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26881         value.
26882         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26883         support.
26884         * config/arc/arc.h (TARGET_DBNZ): Define.
26885         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26886         properly set the tune attribute.
26887         (dbnz): Use TARGET_DBNZ guard.
26888         * config/arc/arc.opt (mtune): Add core3 option.
26890 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26892         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26893         recognize new pic like addresses.
26894         (arc_delegitimize_address): Clean up.
26896 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26898         * config/arc/arc-arches.def: Option mrf16 valid for all
26899         architectures.
26900         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26901         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26902         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26903         * config/arc/arc-tables.opt: Regenerate.
26904         * config/arc/arc.c (arc_conditional_register_usage): Handle
26905         reduced register file case.
26906         (arc_file_start): Set must have build attributes.
26907         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26908         mrf16 option value.
26909         * config/arc/arc.opt (mrf16): Add new option.
26910         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26911         * config/arc/genmultilib.awk: Handle new mrf16 option.
26912         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26913         * config/arc/t-multilib: Regenerate.
26914         * doc/invoke.texi (ARC Options): Document mrf16 option.
26916 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26918         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26919         * config/arc/arc.c (arc_handle_secure_attribute): New function.
26920         (arc_attribute_table): Add 'secure_call' attribute.
26921         (arc_print_operand): Print secure call operand.
26922         (arc_function_ok_for_sibcall): Don't optimize tail calls when
26923         secure.
26924         (arc_is_secure_call_p): New function.  * config/arc/arc.md
26925         (call_i): Add support for sjli instruction.
26926         (call_value_i): Likewise.
26927         * config/arc/constraints.md (Csc): New constraint.
26929 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
26930             John Eric Martin  <John.Martin@emmicro-us.com>
26932         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26933         * config/arc/arc.c (_arc_jli_section): New struct.
26934         (arc_jli_section): New type.
26935         (rc_jli_sections): New static variable.
26936         (arc_handle_jli_attribute): New function.
26937         (arc_attribute_table): Add jli_always and jli_fixed attribute.
26938         (arc_file_end): New function.
26939         (TARGET_ASM_FILE_END): Define.
26940         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26941         (arc_add_jli_section): New function.
26942         (jli_call_scan): Likewise.
26943         (arc_reorg): Call jli_call_scan.
26944         (arc_output_addsi): Remove 'S' from printing asm operand.
26945         (arc_is_jli_call_p): New function.
26946         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26947         operand.
26948         (movhi_insn): Likewise.
26949         (movsi_insn): Likewise.
26950         (movsi_set_cc_insn): Likewise.
26951         (loadqi_update): Likewise.
26952         (load_zeroextendqisi_update): Likewise.
26953         (load_signextendqisi_update): Likewise.
26954         (loadhi_update): Likewise.
26955         (load_zeroextendhisi_update): Likewise.
26956         (load_signextendhisi_update): Likewise.
26957         (loadsi_update): Likewise.
26958         (loadsf_update): Likewise.
26959         (movsicc_insn): Likewise.
26960         (bset_insn): Likewise.
26961         (bxor_insn): Likewise.
26962         (bclr_insn): Likewise.
26963         (bmsk_insn): Likewise.
26964         (bicsi3_insn): Likewise.
26965         (cmpsi_cc_c_insn): Likewise.
26966         (movsi_ne): Likewise.
26967         (movsi_cond_exec): Likewise.
26968         (clrsbsi2): Likewise.
26969         (norm_f): Likewise.
26970         (normw): Likewise.
26971         (swap): Likewise.
26972         (divaw): Likewise.
26973         (flag): Likewise.
26974         (sr): Likewise.
26975         (kflag): Likewise.
26976         (ffs): Likewise.
26977         (ffs_f): Likewise.
26978         (fls): Likewise.
26979         (call_i): Remove 'S' asm letter, add jli instruction.
26980         (call_value_i): Likewise.
26981         * config/arc/arc.op (mjli-always): New option.
26982         * config/arc/constraints.md (Cji): New constraint.
26983         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26984         operand.
26985         (subsf3_fpx): Likewise.
26986         (mulsf3_fpx): Likewise.
26987         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26988         asm operand.
26989         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26990         function attrbutes.
26991         * doc/invoke.texi (ARC): Document mjli-always option.
26993 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
26995         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26996         avoid addition with 0 and use incw and decw where possible.
26998 2018-01-26  Richard Biener  <rguenther@suse.de>
27000         PR tree-optimization/81082
27001         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27002         association if it requires casting to unsigned.
27003         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27004         from fold_plusminus_mult_expr to catch important cases late when
27005         range info is available.
27007 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27009         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27010         * configure.ac (hidden_linkonce): New test.
27011         * configure: Regenerate.
27012         * config.in: Regenerate.
27014 2018-01-26  Julia Koval  <julia.koval@intel.com>
27016         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27017         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27018         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27019         _mm_mask_bitshuffle_epi64_mask): Fix type.
27020         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27021         USI_FTYPE_V4DI_V4DI_USI): Remove.
27022         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27023         __builtin_ia32_vpshufbitqmb256_mask,
27024         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27025         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27026         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27028 2018-01-26  Alan Modra  <amodra@gmail.com>
27030         PR target/84033
27031         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27032         UNSPEC_VBPERMQ.  Sort other unspecs.
27034 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27036         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27038 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27040         PR middle-end/83055
27041         * predict.c (drop_profile): Do not push/pop cfun; update also
27042         node->count.
27043         (handle_missing_profiles): Fix logic looking for zero profiles.
27045 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27047         PR middle-end/83977
27048         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27049         on functions with #pragma omp declare simd or functions with simd
27050         attribute.
27051         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27052         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27053         Remove trailing \n from warning_at calls.
27055 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27057         PR target/84028
27058         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27059         for neutered workers.
27061 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27063         PR target/68467
27064         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27065         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27067 2018-01-24  Jeff Law  <law@redhat.com>
27069         PR target/83994
27070         * i386.c (get_probe_interval): Move to earlier point.
27071         (ix86_compute_frame_layout): If -fstack-clash-protection and
27072         the frame is larger than the probe interval, then use pushes
27073         to save registers rather than reg->mem moves.
27074         (ix86_expand_prologue): Remove conditional for int_registers_saved
27075         assertion.
27077 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27079         PR target/84014
27080         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27081         min/max for never referenced object.
27083 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27085         PR middle-end/83977
27086         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27087         here.
27088         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27089         attributes from DECL_ATTRIBUTES (decl) without affecting
27090         DECL_ATTRIBUTES (current_function_decl).
27091         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27092         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27094 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27096         PR tree-optimization/83979
27097         * fold-const.c (fold_comparison): Use constant_boolean_node
27098         instead of boolean_{true,false}_node.
27100 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27102         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27103         with zero counts.
27105 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27107         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27108         Simplify the clause that sets the length attribute.
27109         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27110         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27111         clause that sets the length attribute.
27112         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27114 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27116         PR target/83589
27117         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27118         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27119         Add strict parameter.
27120         (prevent_branch_around_nothing): Insert dummy insn between branch to
27121         label and label with no ptx insn inbetween.
27122         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27124 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27126         PR target/81352
27127         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27128         for neutered threads in warp.
27129         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27131 2018-01-24  Richard Biener  <rguenther@suse.de>
27133         PR tree-optimization/83176
27134         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27135         operands.
27137 2018-01-24  Richard Biener  <rguenther@suse.de>
27139         PR tree-optimization/82819
27140         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27141         code generating pluses that are no-ops in the target precision.
27143 2018-01-24  Richard Biener  <rguenther@suse.de>
27145         PR middle-end/84000
27146         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27148 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27150         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27151         to merge probabilities.
27152         * predict.c (probably_never_executed): Also mark as cold functions
27153         with global 0 profile and guessed local profile.
27154         * profile-count.c (profile_probability::combine_with_count): New
27155         member function.
27156         * profile-count.h (profile_probability::operator*,
27157         profile_probability::operator*=, profile_probability::operator/,
27158         profile_probability::operator/=): Reduce precision to adjusted
27159         and set value to guessed on contradictory divisions.
27160         (profile_probability::combine_with_freq): Remove.
27161         (profile_probability::combine_wiht_count): Declare.
27162         (profile_count::force_nonzero):: Set to adjusted.
27163         (profile_count::probability_in):: Set quality to adjusted.
27164         * tree-ssa-tail-merge.c (replace_block_by): Use
27165         combine_with_count.
27167 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27168             Jim Wilson  <jimw@sifive.com>
27170         * config/riscv/riscv.c (riscv_stack_boundary): New.
27171         (riscv_option_override): Set riscv_stack_boundary.  Handle
27172         riscv_preferred_stack_boundary_arg.
27173         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27174         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27175         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27176         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27177         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27178         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27180 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27182         PR target/83905
27183         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27184         of struct ix86_frame.
27185         (ix86_expand_epilogue): Likewise.  Add a local variable for
27186         the reg_save_offset field in struct ix86_frame.
27188 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27190         PR tree-optimization/82604
27191         * tree-loop-distribution.c (enum partition_kind): New enum item
27192         PKIND_PARTIAL_MEMSET.
27193         (partition_builtin_p): Support above new enum item.
27194         (generate_code_for_partition): Ditto.
27195         (compute_access_range): Differentiate cases that equality can be
27196         proven at all loops, the innermost loops or no loops.
27197         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27198         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27199         (finalize_partitions, distribute_loop): Don't fuse partition of
27200         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27201         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27202         parloop is enabled.
27204 2018-01-23  Martin Liska  <mliska@suse.cz>
27206         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27207         order to ignore the predictor.
27208         (PRED_POLYMORPHIC_CALL): Likewise.
27209         (PRED_RECURSIVE_CALL): Likewise.
27211 2018-01-23  Martin Liska  <mliska@suse.cz>
27213         * tree-profile.c (tree_profiling): Print function header to
27214         aware reader which function we are working on.
27215         * value-prof.c (gimple_find_values_to_profile): Do not print
27216         not interesting value histograms.
27218 2018-01-23  Martin Liska  <mliska@suse.cz>
27220         * profile-count.h (enum profile_quality): Add
27221         profile_uninitialized as the first value. Do not number values
27222         as they are zero based.
27223         (profile_count::verify): Update sanity check.
27224         (profile_probability::verify): Likewise.
27226 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27228         * doc/invoke.texi (ffor-scope): Deprecate.
27230 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27232         PR tree-optimization/83510
27233         * domwalk.c (set_all_edges_as_executable): New function.
27234         (dom_walker::dom_walker): Convert bool param
27235         "skip_unreachable_blocks" to enum reachability.  Move setup of
27236         edge flags to set_all_edges_as_executable and only do it when
27237         reachability is REACHABLE_BLOCKS.
27238         * domwalk.h (enum dom_walker::reachability): New enum.
27239         (dom_walker::dom_walker): Convert bool param
27240         "skip_unreachable_blocks" to enum reachability.
27241         (set_all_edges_as_executable): New decl.
27242         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27243         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27244         "reachability".
27245         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27246         but converting true to REACHABLE_BLOCKS.
27247         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27248         * tree-vrp.c
27249         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27250         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27251         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27252         REACHABLE_BLOCKS.
27253         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27254         if check_all_array_refs will be called.
27256 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27258         * tree.c (selftest::test_location_wrappers): Add more test
27259         coverage.
27261 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27263         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27264         (selftest::test_bit_in_range): Likewise.
27266 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27268         PR testsuite/83888
27269         * doc/sourcebuild.texi (vect_float): Say that the selector
27270         only describes the situation when -funsafe-math-optimizations is on.
27271         (vect_float_strict): Document.
27273 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27275         PR tree-optimization/83965
27276         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27277         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27278         instead of checking only for a reduction.
27279         (vect_recog_widen_sum_pattern): Likewise.
27281 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27283         * predict.c (probably_never_executed): Only use precise profile info.
27284         (compute_function_frequency): Skip after inlining hack since we now
27285         have quality checking.
27287 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27289         * profile-count.h (profile_probability::very_unlikely,
27290         profile_probability::unlikely, profile_probability::even): Set
27291         precision to guessed.
27293 2018-01-23  Richard Biener  <rguenther@suse.de>
27295         PR tree-optimization/83963
27296         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27297         Properly terminate dominator walk when crossing the exit edge not
27298         when visiting its source block.
27300 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27302         PR c++/83918
27303         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27304         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27306 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27308         PR tree-optimization/83957
27309         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27310         semicolon after for body surrounded by braces.
27312         PR tree-optimization/83081
27313         * profile-count.h (profile_probability::split): New method.
27314         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27315         Use profile_probability::split.
27316         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27317         when splitting a single conditional jump into 2.
27319 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27321         PR tree-optimization/69452
27322         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27323         decl.
27325 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27327         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27328         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27329         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27331 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27333         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27334         declaration.
27335         * config/rl78/rl78.md (movdi): New define_expand.
27336         * config/rl78/rl78.c (rl78_split_movdi): New function.
27338 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27340         PR target/83862
27341         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27342         no longer used.
27343         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27344         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27345         128-bit to produce an UNSPEC move to get the double word with the
27346         signbit and then a shift directly to do signbit.
27347         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27348         implementation with a new version that just does either a direct
27349         move or a regular move.  Move memory interface to separate insns.
27350         Move insns so they are next to the expander.
27351         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27352         with signbit move.  Split big and little endian case.
27353         (signbit<mode>2_dm_mem_le): Likewise.
27354         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27355         (signbit<mode>2_dm2): Likewise.
27357 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27359         * config/rl78/rl78.md (anddi3): New define_expand.
27361 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27363         * config/rl78/rl78.md (umindi3): New define_expand.
27365 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27367         * config/rl78/rl78.md (smindi3): New define_expand.
27369 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27371         * config/rl78/rl78.md (smaxdi3): New define_expand.
27373 2018-01-22  Carl Love  <cel@us.ibm.com>
27375         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27376         LVX_V1TI): Add macro expansion.
27377         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27378         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27379         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27380         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27381         Change check to determine if the instruction is a byte reversing
27382         entry.  Fix typo in comment.
27383         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27384         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27385         Add def_builtin calls for new builtins.
27386         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27387         Add define_insn expansion.
27389 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27391         * config/rl78/rl78.md (umaxdi3): New define_expand.
27393 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27395         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27396         for non-QImode registers.
27398 2018-01-22  Richard Biener  <rguenther@suse.de>
27400         PR tree-optimization/83963
27401         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27402         including the loop exit block.
27403         (scop_detection::merge_sese): Likewise.
27404         (scop_detection::add_scop): Do it here instead.
27406 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27408         * doc/sourcebuild.texi (arm_softfloat): Document.
27410 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27412         PR gcc/77734
27413         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27414         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27415         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27417 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27418             David Edelsohn  <dje.gcc@gmail.com>
27420         PR target/83946
27421         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27422         Change "crset eq" to "crset 2".
27423         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27424         (*call_indirect_aix<mode>_nospec): Likewise.
27425         (*call_value_indirect_aix<mode>_nospec): Likewise.
27426         (*call_indirect_elfv2<mode>_nospec): Likewise.
27427         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27428         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27429         change assembly output from . to $.
27430         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27431         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27432         (*tablejump<mode>_internal1_nospec): Likewise.
27434 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27436         PR target/80870
27437         * config/sh/sh_optimize_sett_clrt.cc:
27438         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27440 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27442         PR tree-optimization/83940
27443         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27444         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27445         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27446         use it to pass back the definition type.
27447         (vect_check_store_rhs): Likewise rhs_dt_out.
27448         (vect_build_gather_load_calls): Add a mask_dt argument and use
27449         it instead of a call to vect_is_simple_use.
27450         (vectorizable_store): Update calls to vect_check_load_store_mask
27451         and vect_check_store_rhs.  Use the dt returned by the latter instead
27452         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27453         instead of calls to vect_is_simple_use.  Pass the scalar rather
27454         than the vector operand to vect_is_simple_use when handling
27455         second and subsequent copies of an rhs value.
27456         (vectorizable_load): Update calls to vect_check_load_store_mask
27457         and vect_build_gather_load_calls.  Use the cached mask_dt and
27458         gs_info.offset_dt instead of calls to vect_is_simple_use.
27460 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27462         PR middle-end/83945
27463         * tree-emutls.c: Include gimplify.h.
27464         (lower_emutls_2): New function.
27465         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27466         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27467         it before further processing.
27469         PR target/83930
27470         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27471         UINTVAL (trueop1) instead of INTVAL (op1).
27473 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27475         PR debug/81570
27476         PR debug/83728
27477         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27478         INCOMING_FRAME_SP_OFFSET if not defined.
27479         (scan_trace): Add ENTRY argument.  If true and
27480         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27481         emit a note to adjust the CFA offset.
27482         (create_cfi_notes): Adjust scan_trace callers.
27483         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27484         INCOMING_FRAME_SP_OFFSET in the CIE.
27485         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27486         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27487         Likewise.
27488         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27489         * doc/tm.texi: Regenerated.
27491 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27493         PR rtl-optimization/83147
27494         * lra-constraints.c (remove_inheritance_pseudos): Use
27495         lra_substitute_pseudo_within_insn.
27497 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27498             Cesar Philippidis  <cesar@codesourcery.com>
27500         PR target/83920
27501         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27503 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27505         PR target/83790
27506         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27507         spaces for function labels.
27509 2018-01-19  Martin Liska  <mliska@suse.cz>
27511         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27512         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27513         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27514         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27515         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27516         (PRED_CONST_RETURN): Change from 69 to 65.
27517         (PRED_NULL_RETURN): Change from 91 to 71.
27518         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27519         (PRED_LOOP_GUARD): Change from 66 to 73.
27521 2018-01-19  Martin Liska  <mliska@suse.cz>
27523         * predict.c (predict_insn_def): Add new assert.
27524         (struct branch_predictor): Change type to signed integer.
27525         (test_prediction_value_range): Amend test to cover
27526         PROB_UNINITIALIZED.
27527         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27528         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27529         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27530         (PRED_LOOP_IV_COMPARE): Likewise.
27531         * predict.h (PROB_UNINITIALIZED): Define new constant.
27533 2018-01-19  Martin Liska  <mliska@suse.cz>
27535         * predict.c (dump_prediction): Add new format for
27536         analyze_brprob.py script which is enabled with -details
27537         suboption.
27538         * profile-count.h (precise_p): New function.
27540 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27542         PR tree-optimization/83922
27543         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27544         there are no statements that need masking.
27545         (vect_active_double_reduction_p): New function.
27546         (vect_analyze_loop_operations): Use it when handling phis that
27547         are not in the loop header.
27549 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27551         PR tree-optimization/83914
27552         * tree-vect-loop.c (vectorizable_induction): Don't convert
27553         init_expr or apply the peeling adjustment for inductions
27554         that are nested within the vectorized loop.
27556 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27558         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27559         instead of NEG.
27561 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27563         PR sanitizer/81715
27564         PR testsuite/83882
27565         * function.h (gimplify_parameters): Add gimple_seq * argument.
27566         * function.c: Include gimple.h and options.h.
27567         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27568         for the added local temporaries if needed.
27569         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27570         if there are any parameter cleanups, wrap whole body into a
27571         try/finally with the cleanups.
27573 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27575         PR target/82964
27576         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27577         Use GET_MODE_CLASS for scalar floating point.
27579 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27581         PR ipa/82256
27582         patch by PaX Team
27583         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27584         Fix call of call_cgraph_insertion_hooks.
27586 2018-01-18  Martin Sebor  <msebor@redhat.com>
27588         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27590 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27592         PR ipa/83619
27593         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27594         frequencies.
27596 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27598         PR other/70268
27599         * common.opt: (-ffile-prefix-map): New option.
27600         * opts.c (common_handle_option): Defer it.
27601         * opts-global.c (handle_common_deferred_options): Handle it.
27602         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27603         * file-prefix-map.h: New file.
27604         (remap_debug_filename, add_debug_prefix_map): ...here.
27605         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27606         * final.c (debug_prefix_map, add_debug_prefix_map
27607         remap_debug_filename): Move to...
27608         * file-prefix-map.c: New file.
27609         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27610         generalize, get rid of alloca(), use strrchr() instead of strchr().
27611         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27612         Implement in terms of add_prefix_map().
27613         (remap_macro_filename, remap_debug_filename): Implement in term of
27614         remap_filename().
27615         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27616         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27617         * dbxout.c: Include file-prefix-map.h.
27618         * varasm.c: Likewise.
27619         * vmsdbgout.c: Likewise.
27620         * xcoffout.c: Likewise.
27621         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27622         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27623         * doc/invoke.texi (-ffile-prefix-map): Document.
27624         (-fdebug-prefix-map): Update description.
27626 2018-01-18  Martin Liska  <mliska@suse.cz>
27628         * config/i386/i386.c (indirect_thunk_name): Document that also
27629         lfence is emitted.
27630         (output_indirect_thunk): Document why both instructions
27631         (pause and lfence) are generated.
27633 2018-01-18  Richard Biener  <rguenther@suse.de>
27635         PR tree-optimization/83887
27636         * graphite-scop-detection.c
27637         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27638         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27639         (scop_detection::merge_sese): Re-implement with a flood-fill
27640         algorithm that properly finds a SESE region if it exists.
27642 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27644         PR c/61240
27645         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27646         pointer_diff optimizations use view_convert instead of convert.
27648 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27650         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27651         Generate different code for -mno-speculate-indirect-jumps.
27652         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27653         (*call_indirect_aix<mode>): Disable for
27654         -mno-speculate-indirect-jumps.
27655         (*call_indirect_aix<mode>_nospec): New define_insn.
27656         (*call_value_indirect_aix<mode>): Disable for
27657         -mno-speculate-indirect-jumps.
27658         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27659         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27660         -mno-speculate-indirect-jumps.
27661         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27663 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27665         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27666         long double type, set the flags for noting the default long double
27667         type, even if we don't pass or return a long double type.
27669 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27671         PR ipa/83051
27672         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27673         failure.
27675 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27677         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27678         support for merge[hl].
27679         (fold_mergehl_helper): New helper function.
27680         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27681         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27682         (altivec_vmrglw_direct): Add xxmrglw insn.
27684 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27686         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27687         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27689 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27691         PR lto/83121
27692         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27693         call the lto_location_cache before reading the
27694         DECL_SOURCE_LOCATION of the types.
27696 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27697             Richard Sandiford  <richard.sandiford@linaro.org>
27699         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27700         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27701         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27702         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27703         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27704         Add declaration.
27705         * config/aarch64/constraints.md (aarch64_movti_operand):
27706         Limit immediates.
27707         * config/aarch64/predicates.md (Uti): Add new constraint.
27709 2018-01-17  Carl Love  <cel@us.ibm.com>
27711         * config/rs6000/vsx.md (define_expand xl_len_r,
27712         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27713         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27714         lxvll.
27715         (define_expand, define_insn): Move the shift left from  the
27716         define_insn to the define_expand for lxvl and stxvl instructions.
27717         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27718         and XL_LEN_R definitions to PURE.
27720 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27722         * config/i386/i386.c (indirect_thunk_name): Declare regno
27723         as unsigned int.  Compare regno with INVALID_REGNUM.
27724         (output_indirect_thunk): Ditto.
27725         (output_indirect_thunk_function): Ditto.
27726         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27727         in the call to output_indirect_thunk_function.
27729 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27731         PR middle-end/83884
27732         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27733         rather than the size of inner_type to determine the stack slot size
27734         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27736 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27738         PR target/83546
27739         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27740         to PTA_SILVERMONT.
27742 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27744         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27745         endian Linux systems to optionally enable multilibs for selecting
27746         the long double type if the user configured an explicit type.
27747         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27748         have no long double multilibs if not defined.
27749         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27750         warn if the user used -mabi={ieee,ibm}longdouble and we built
27751         multilibs for long double.
27752         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27753         appropriate multilib option.
27754         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27755         multilib options.
27756         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27757         for building long double multilibs.
27758         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27760 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27762         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27763         copies.
27765         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27766         64 bits.
27767         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27768         128 bits.
27770         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27771         variables.
27773         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27774         return value.
27776 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27778         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27779         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27781 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27783         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27784         different rtl trees depending on TARGET_64BIT.
27785         (rs6000_gen_lvx): Likewise.
27787 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27789         * config/visium/visium.md (nop): Tweak comment.
27790         (hazard_nop): Likewise.
27792 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27794         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27795         -mspeculate-indirect-jumps.
27796         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27797         for -mno-speculate-indirect-jumps.
27798         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27799         (*call_value_indirect_elfv2<mode>): Disable for
27800         -mno-speculate-indirect-jumps.
27801         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27802         (indirect_jump): Emit different RTL for
27803         -mno-speculate-indirect-jumps.
27804         (*indirect_jump<mode>): Disable for
27805         -mno-speculate-indirect-jumps.
27806         (*indirect_jump<mode>_nospec): New define_insn.
27807         (tablejump): Emit different RTL for
27808         -mno-speculate-indirect-jumps.
27809         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27810         (tablejumpsi_nospec): New define_expand.
27811         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27812         (tablejumpdi_nospec): New define_expand.
27813         (*tablejump<mode>_internal1): Disable for
27814         -mno-speculate-indirect-jumps.
27815         (*tablejump<mode>_internal1_nospec): New define_insn.
27816         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27817         option.
27819 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27821         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27822         callers updated.
27824 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27825             Richard Biener  <rguenth@suse.de>
27827         PR libgomp/83590
27828         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27829         return early, inline manually is_gimple_sizepos.  Make sure if we
27830         call gimplify_expr we don't end up with a gimple constant.
27831         * tree.c (variably_modified_type_p): Don't return true for
27832         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27833         * gimplify.h (is_gimple_sizepos): Remove.
27835 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27837         PR tree-optimization/83857
27838         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27839         vectorizable_live_operation for pure SLP statements.
27840         (vectorizable_live_operation): Handle PHIs.
27842 2018-01-16  Richard Biener  <rguenther@suse.de>
27844         PR tree-optimization/83867
27845         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27846         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27848 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27850         PR c/83844
27851         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27852         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27853         If off is not INTEGER_CST, issue a may not be aligned warning
27854         rather than isn't aligned.  Use isn%'t rather than isn't.
27855         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27856         into MULT_EXPR.
27857         <case MULT_EXPR>: Improve the case when bottom and one of the
27858         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27859         operand, in that case check if the other operand is multiple of
27860         bottom divided by the INTEGER_CST operand.
27862 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27864         PR target/83858
27865         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27866         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27867         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27868         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27869         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27870         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27871         (pa_function_arg_size): New function.
27873 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27875         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27876         in a separate statement.
27878 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27880         PR tree-optimization/83847
27881         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27882         group gathers and scatters.
27884 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27886         PR rtl-optimization/86620
27887         * params.def (max-sched-ready-insns): Bump minimum value to 1.
27889         PR rtl-optimization/83213
27890         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27891         to last if both are JUMP_INSNs.
27893         PR tree-optimization/83843
27894         * gimple-ssa-store-merging.c
27895         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27896         store_immediate_info for bswap/nop orig_stores.
27898 2018-01-15  Andrew Waterman  <andrew@sifive.com>
27900         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27901         !TARGET_MUL.
27902         <UDIV>: Increase cost if !TARGET_DIV.
27904 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
27906         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27907         (define_attr "cr_logical_3op"): New.
27908         (cceq_ior_compare): Adjust.
27909         (cceq_ior_compare_complement): Adjust.
27910         (*cceq_rev_compare): Adjust.
27911         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27912         (is_cracked_insn): Adjust.
27913         (insn_must_be_first_in_group): Adjust.
27914         * config/rs6000/40x.md: Adjust.
27915         * config/rs6000/440.md: Adjust.
27916         * config/rs6000/476.md: Adjust.
27917         * config/rs6000/601.md: Adjust.
27918         * config/rs6000/603.md: Adjust.
27919         * config/rs6000/6xx.md: Adjust.
27920         * config/rs6000/7450.md: Adjust.
27921         * config/rs6000/7xx.md: Adjust.
27922         * config/rs6000/8540.md: Adjust.
27923         * config/rs6000/cell.md: Adjust.
27924         * config/rs6000/e300c2c3.md: Adjust.
27925         * config/rs6000/e500mc.md: Adjust.
27926         * config/rs6000/e500mc64.md: Adjust.
27927         * config/rs6000/e5500.md: Adjust.
27928         * config/rs6000/e6500.md: Adjust.
27929         * config/rs6000/mpc.md: Adjust.
27930         * config/rs6000/power4.md: Adjust.
27931         * config/rs6000/power5.md: Adjust.
27932         * config/rs6000/power6.md: Adjust.
27933         * config/rs6000/power7.md: Adjust.
27934         * config/rs6000/power8.md: Adjust.
27935         * config/rs6000/power9.md: Adjust.
27936         * config/rs6000/rs64.md: Adjust.
27937         * config/rs6000/titan.md: Adjust.
27939 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27941         * config/i386/predicates.md (indirect_branch_operand): Rewrite
27942         ix86_indirect_branch_register logic.
27944 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27946         * config/i386/constraints.md (Bs): Update
27947         ix86_indirect_branch_register check.  Don't check
27948         ix86_indirect_branch_register with GOT_memory_operand.
27949         (Bw): Likewise.
27950         * config/i386/predicates.md (GOT_memory_operand): Don't check
27951         ix86_indirect_branch_register here.
27952         (GOT32_symbol_operand): Likewise.
27954 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27956         * config/i386/predicates.md (constant_call_address_operand):
27957         Rewrite ix86_indirect_branch_register logic.
27958         (sibcall_insn_operand): Likewise.
27960 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27962         * config/i386/constraints.md (Bs): Replace
27963         ix86_indirect_branch_thunk_register with
27964         ix86_indirect_branch_register.
27965         (Bw): Likewise.
27966         * config/i386/i386.md (indirect_jump): Likewise.
27967         (tablejump): Likewise.
27968         (*sibcall_memory): Likewise.
27969         (*sibcall_value_memory): Likewise.
27970         Peepholes of indirect call and jump via memory: Likewise.
27971         * config/i386/i386.opt: Likewise.
27972         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27973         (GOT_memory_operand): Likewise.
27974         (call_insn_operand): Likewise.
27975         (sibcall_insn_operand): Likewise.
27976         (GOT32_symbol_operand): Likewise.
27978 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
27980         PR middle-end/83837
27981         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27982         type rather than type addr's type points to.
27983         (expand_omp_atomic_mutex): Likewise.
27984         (expand_omp_atomic): Likewise.
27986 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
27988         PR target/83839
27989         * config/i386/i386.c (output_indirect_thunk_function): Use
27990         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27991         for  __x86_return_thunk.
27993 2018-01-15  Richard Biener  <rguenther@suse.de>
27995         PR middle-end/83850
27996         * expmed.c (extract_bit_field_1): Fix typo.
27998 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28000         PR target/83687
28001         * config/arm/iterators.md (VF): New mode iterator.
28002         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28003         Remove integer-related logic from pattern.
28004         (neon_vabd<mode>_3): Likewise.
28006 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28008         PR middle-end/82694
28009         * common.opt (fstrict-overflow): No longer an alias.
28010         (fwrapv-pointer): New option.
28011         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28012         also for pointer types based on flag_wrapv_pointer.
28013         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28014         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28015         opts->x_flag_wrapv got set.
28016         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28017         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28018         POINTER_TYPE_OVERFLOW_UNDEFINED.
28019         * match.pd: Likewise in address comparison pattern.
28020         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28022 2018-01-15  Richard Biener  <rguenther@suse.de>
28024         PR lto/83804
28025         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28026         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28027         Reset type names to their identifier if their TYPE_DECL doesn't
28028         have linkage (and thus is used for ODR and devirt).
28029         (save_debug_info_for_decl): Remove.
28030         (save_debug_info_for_type): Likewise.
28031         (add_tree_to_fld_list): Adjust.
28032         * tree-pretty-print.c (dump_generic_node): Make dumping of
28033         type names more robust.
28035 2018-01-15  Richard Biener  <rguenther@suse.de>
28037         * BASE-VER: Bump to 8.0.1.
28039 2018-01-14  Martin Sebor  <msebor@redhat.com>
28041         PR other/83508
28042         * builtins.c (check_access): Avoid warning when the no-warning bit
28043         is set.
28045 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28047         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28048         * ira-color (allocno_hard_regs_compare): Likewise.
28050 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28052         PR target/83013
28053         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28054         Use .pushsection/.popsection.
28056 2018-01-14  Martin Sebor  <msebor@redhat.com>
28058         PR c++/81327
28059         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28061 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28063         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28064         entry from extra_headers.
28065         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28066         extra_headers, make the list bitwise identical to the i?86-*-* one.
28068 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28070         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28071         -mcmodel=large with -mindirect-branch=thunk,
28072         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28073         -mfunction-return=thunk-extern.
28074         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28075         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28076         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28078 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28080         * config/i386/i386.c (print_reg): Print the name of the full
28081         integer register without '%'.
28082         (ix86_print_operand): Handle 'V'.
28083         * doc/extend.texi: Document 'V' modifier.
28085 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28087         * config/i386/constraints.md (Bs): Disallow memory operand for
28088         -mindirect-branch-register.
28089         (Bw): Likewise.
28090         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28091         (GOT_memory_operand): Likewise.
28092         (call_insn_operand): Likewise.
28093         (sibcall_insn_operand): Likewise.
28094         (GOT32_symbol_operand): Likewise.
28095         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28096         for -mindirect-branch-register.
28097         (tablejump): Likewise.
28098         (*sibcall_memory): Likewise.
28099         (*sibcall_value_memory): Likewise.
28100         Disallow peepholes of indirect call and jump via memory for
28101         -mindirect-branch-register.
28102         (*call_pop): Replace m with Bw.
28103         (*call_value_pop): Likewise.
28104         (*sibcall_pop_memory): Replace m with Bs.
28105         * config/i386/i386.opt (mindirect-branch-register): New option.
28106         * doc/invoke.texi: Document -mindirect-branch-register option.
28108 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28110         * config/i386/i386-protos.h (ix86_output_function_return): New.
28111         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28112         set function_return_type.
28113         (indirect_thunk_name): Add ret_p to indicate thunk for function
28114         return.
28115         (output_indirect_thunk_function): Pass false to
28116         indirect_thunk_name.
28117         (ix86_output_indirect_branch_via_reg): Likewise.
28118         (ix86_output_indirect_branch_via_push): Likewise.
28119         (output_indirect_thunk_function): Create alias for function
28120         return thunk if regno < 0.
28121         (ix86_output_function_return): New function.
28122         (ix86_handle_fndecl_attribute): Handle function_return.
28123         (ix86_attribute_table): Add function_return.
28124         * config/i386/i386.h (machine_function): Add
28125         function_return_type.
28126         * config/i386/i386.md (simple_return_internal): Use
28127         ix86_output_function_return.
28128         (simple_return_internal_long): Likewise.
28129         * config/i386/i386.opt (mfunction-return=): New option.
28130         (indirect_branch): Mention -mfunction-return=.
28131         * doc/extend.texi: Document function_return function attribute.
28132         * doc/invoke.texi: Document -mfunction-return= option.
28134 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28136         * config/i386/i386-opts.h (indirect_branch): New.
28137         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28138         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28139         with local indirect jump when converting indirect call and jump.
28140         (ix86_set_indirect_branch_type): New.
28141         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28142         (indirectlabelno): New.
28143         (indirect_thunk_needed): Likewise.
28144         (indirect_thunk_bnd_needed): Likewise.
28145         (indirect_thunks_used): Likewise.
28146         (indirect_thunks_bnd_used): Likewise.
28147         (INDIRECT_LABEL): Likewise.
28148         (indirect_thunk_name): Likewise.
28149         (output_indirect_thunk): Likewise.
28150         (output_indirect_thunk_function): Likewise.
28151         (ix86_output_indirect_branch_via_reg): Likewise.
28152         (ix86_output_indirect_branch_via_push): Likewise.
28153         (ix86_output_indirect_branch): Likewise.
28154         (ix86_output_indirect_jmp): Likewise.
28155         (ix86_code_end): Call output_indirect_thunk_function if needed.
28156         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28157         needed.
28158         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28159         (ix86_attribute_table): Add indirect_branch.
28160         * config/i386/i386.h (machine_function): Add indirect_branch_type
28161         and has_local_indirect_jump.
28162         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28163         to true.
28164         (tablejump): Likewise.
28165         (*indirect_jump): Use ix86_output_indirect_jmp.
28166         (*tablejump_1): Likewise.
28167         (simple_return_indirect_internal): Likewise.
28168         * config/i386/i386.opt (mindirect-branch=): New option.
28169         (indirect_branch): New.
28170         (keep): Likewise.
28171         (thunk): Likewise.
28172         (thunk-inline): Likewise.
28173         (thunk-extern): Likewise.
28174         * doc/extend.texi: Document indirect_branch function attribute.
28175         * doc/invoke.texi: Document -mindirect-branch= option.
28177 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28179         PR ipa/83051
28180         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28182 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28184         * ipa-inline.c (want_inline_small_function_p): Return false if
28185         inlining has already failed with CIF_FINAL_ERROR.
28186         (update_caller_keys): Call want_inline_small_function_p before
28187         can_inline_edge_p.
28188         (update_callee_keys): Likewise.
28190 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28192         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28193         New function.
28194         (rs6000_quadword_masked_address_p): Likewise.
28195         (quad_aligned_load_p): Likewise.
28196         (quad_aligned_store_p): Likewise.
28197         (const_load_sequence_p): Add comment to describe the outer-most loop.
28198         (mimic_memory_attributes_and_flags): New function.
28199         (rs6000_gen_stvx): Likewise.
28200         (replace_swapped_aligned_store): Likewise.
28201         (rs6000_gen_lvx): Likewise.
28202         (replace_swapped_aligned_load): Likewise.
28203         (replace_swapped_load_constant): Capitalize argument name in
28204         comment describing this function.
28205         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28206         and stores that access quad-word aligned addresses and replace
28207         with stvx or lvx instructions when appropriate.
28208         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28209         New function prototype.
28210         (rs6000_quadword_masked_address_p): Likewise.
28211         (rs6000_gen_lvx): Likewise.
28212         (rs6000_gen_stvx): Likewise.
28213         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28214         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28215         when memory address is aligned.
28216         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28217         this split to select lvx instruction when memory address is aligned.
28218         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28219         instruction when memory address is aligned.
28220         (*vsx_le_perm_load_v16qi): Likewise.
28221         (four unnamed splitters): Modify to select the stvx instruction
28222         when memory is aligned.
28224 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28226         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28227         which appears in the queue multiple times.
28229 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28230             Alan Hayward  <alan.hayward@arm.com>
28231             David Sherwood  <david.sherwood@arm.com>
28233         * tree-vectorizer.h (vec_lower_bound): New structure.
28234         (_loop_vec_info): Add check_nonzero and lower_bounds.
28235         (LOOP_VINFO_CHECK_NONZERO): New macro.
28236         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28237         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28238         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28239         fields.  Make seg_len the distance travelled, not including the
28240         access size.
28241         (dr_direction_indicator): Declare.
28242         (dr_zero_step_indicator): Likewise.
28243         (dr_known_forward_stride_p): Likewise.
28244         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28245         tree-ssanames.h.
28246         (runtime_alias_check_p): Allow runtime alias checks with
28247         variable strides.
28248         (operator ==): Compare access_size and align.
28249         (prune_runtime_alias_test_list): Rework for new distinction between
28250         the access_size and seg_len.
28251         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28252         segment lengths.
28253         (get_segment_min_max): New function.
28254         (create_intersect_range_checks): Use it.
28255         (dr_step_indicator): New function.
28256         (dr_direction_indicator): Likewise.
28257         (dr_zero_step_indicator): Likewise.
28258         (dr_known_forward_stride_p): Likewise.
28259         * tree-loop-distribution.c (data_ref_segment_size): Return
28260         DR_STEP * (niters - 1).
28261         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28262         constructor.
28263         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28264         (vect_preserves_scalar_order_p): New function, split out from...
28265         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28266         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28267         (vect_vfa_access_size): New function.
28268         (vect_vfa_align): Likewise.
28269         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28270         (dump_lower_bound): New function.
28271         (vect_check_lower_bound): Likewise.
28272         (vect_small_gap_p): Likewise.
28273         (vectorizable_with_step_bound_p): Likewise.
28274         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28275         depencies if the vectorization factor is 1.  Convert the checks
28276         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28277         a bunds check for variable steps if the minimum required step is
28278         relatively small. Update calls to the dr_with_seg_len
28279         constructor and to vect_compile_time_alias.
28280         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28281         function.
28282         (vect_loop_versioning): Call it.
28283         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28284         when retrying.
28285         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28287 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28288             Alan Hayward  <alan.hayward@arm.com>
28289             David Sherwood  <david.sherwood@arm.com>
28291         * doc/sourcebuild.texi (vect_scatter_store): Document.
28292         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28293         optabs.
28294         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28295         Document.
28296         * genopinit.c (main): Add supports_vec_scatter_store and
28297         supports_vec_scatter_store_cached to target_optabs.
28298         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28299         IFN_MASK_SCATTER_STORE.
28300         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28301         functions.
28302         * internal-fn.h (internal_store_fn_p): Declare.
28303         (internal_fn_stored_value_index): Likewise.
28304         * internal-fn.c (scatter_store_direct): New macro.
28305         (expand_scatter_store_optab_fn): New function.
28306         (direct_scatter_store_optab_supported_p): New macro.
28307         (internal_store_fn_p): New function.
28308         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28309         IFN_MASK_SCATTER_STORE.
28310         (internal_fn_mask_index): Likewise.
28311         (internal_fn_stored_value_index): New function.
28312         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28313         for scatter stores.
28314         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28315         * optabs-query.c (supports_vec_scatter_store_p): New function.
28316         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28317         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28318         true for scatter stores.
28319         (vect_gather_scatter_fn_p): Handle scatter stores too.
28320         (vect_check_gather_scatter): Consider using scatter stores if
28321         supports_vec_scatter_store_p.
28322         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28323         scatter stores too.
28324         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28325         internal_fn_stored_value_index.
28326         (check_load_store_masking): Handle scatter stores too.
28327         (vect_get_store_rhs): Make public.
28328         (vectorizable_call): Use internal_store_fn_p.
28329         (vectorizable_store): Handle scatter store internal functions.
28330         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28331         when deciding whether the end of the group has been reached.
28332         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28333         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28334         (mask_scatter_store<mode>): New insns.
28336 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28337             Alan Hayward  <alan.hayward@arm.com>
28338             David Sherwood  <david.sherwood@arm.com>
28340         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28341         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28342         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28343         function.
28344         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28345         Use vect_truncate_gather_scatter_offset if we can't treat the
28346         operation as a normal gather load or scatter store.
28347         (get_group_load_store_type): Take the gather_scatter_info
28348         as argument.  Try using a gather load or scatter store for
28349         single-element groups.
28350         (get_load_store_type): Update calls to get_group_load_store_type
28351         and vect_use_strided_gather_scatters_p.
28353 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28354             Alan Hayward  <alan.hayward@arm.com>
28355             David Sherwood  <david.sherwood@arm.com>
28357         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28358         optional tree argument.
28359         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28360         null target hooks.
28361         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28362         but continue to use the current value as a fallback.
28363         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28364         to compare the updates.
28365         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28366         (get_load_store_type): Use it when handling a strided access.
28367         (vect_get_strided_load_store_ops): New function.
28368         (vect_get_data_ptr_increment): Likewise.
28369         (vectorizable_load): Handle strided gather loads.  Always pass
28370         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28372 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28373             Alan Hayward  <alan.hayward@arm.com>
28374             David Sherwood  <david.sherwood@arm.com>
28376         * doc/md.texi (gather_load@var{m}): Document.
28377         (mask_gather_load@var{m}): Likewise.
28378         * genopinit.c (main): Add supports_vec_gather_load and
28379         supports_vec_gather_load_cached to target_optabs.
28380         * optabs-tree.c (init_tree_optimization_optabs): Use
28381         ggc_cleared_alloc to allocate target_optabs.
28382         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28383         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28384         functions.
28385         * internal-fn.h (internal_load_fn_p): Declare.
28386         (internal_gather_scatter_fn_p): Likewise.
28387         (internal_fn_mask_index): Likewise.
28388         (internal_gather_scatter_fn_supported_p): Likewise.
28389         * internal-fn.c (gather_load_direct): New macro.
28390         (expand_gather_load_optab_fn): New function.
28391         (direct_gather_load_optab_supported_p): New macro.
28392         (direct_internal_fn_optab): New function.
28393         (internal_load_fn_p): Likewise.
28394         (internal_gather_scatter_fn_p): Likewise.
28395         (internal_fn_mask_index): Likewise.
28396         (internal_gather_scatter_fn_supported_p): Likewise.
28397         * optabs-query.c (supports_at_least_one_mode_p): New function.
28398         (supports_vec_gather_load_p): Likewise.
28399         * optabs-query.h (supports_vec_gather_load_p): Declare.
28400         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28401         and memory_type field.
28402         (NUM_PATTERNS): Bump to 15.
28403         * tree-vect-data-refs.c: Include internal-fn.h.
28404         (vect_gather_scatter_fn_p): New function.
28405         (vect_describe_gather_scatter_call): Likewise.
28406         (vect_check_gather_scatter): Try using internal functions for
28407         gather loads.  Recognize existing calls to a gather load function.
28408         (vect_analyze_data_refs): Consider using gather loads if
28409         supports_vec_gather_load_p.
28410         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28411         (vect_get_gather_scatter_offset_type): Likewise.
28412         (vect_convert_mask_for_vectype): Likewise.
28413         (vect_add_conversion_to_patterm): Likewise.
28414         (vect_try_gather_scatter_pattern): Likewise.
28415         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28416         (vect_vect_recog_func_ptrs): Add it.
28417         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28418         internal_fn_mask_index and internal_gather_scatter_fn_p.
28419         (check_load_store_masking): Take the gather_scatter_info as an
28420         argument and handle gather loads.
28421         (vect_get_gather_scatter_ops): New function.
28422         (vectorizable_call): Check internal_load_fn_p.
28423         (vectorizable_load): Likewise.  Handle gather load internal
28424         functions.
28425         (vectorizable_store): Update call to check_load_store_masking.
28426         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28427         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28428         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28429         (aarch64_gather_scale_operand_d): New predicates.
28430         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28431         (mask_gather_load<mode>): New insns.
28433 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28434             Alan Hayward  <alan.hayward@arm.com>
28435             David Sherwood  <david.sherwood@arm.com>
28437         * optabs.def (fold_left_plus_optab): New optab.
28438         * doc/md.texi (fold_left_plus_@var{m}): Document.
28439         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28440         * internal-fn.c (fold_left_direct): Define.
28441         (expand_fold_left_optab_fn): Likewise.
28442         (direct_fold_left_optab_supported_p): Likewise.
28443         * fold-const-call.c (fold_const_fold_left): New function.
28444         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28445         * tree-parloops.c (valid_reduction_p): New function.
28446         (gather_scalar_reductions): Use it.
28447         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28448         (vect_finish_replace_stmt): Declare.
28449         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28450         (needs_fold_left_reduction_p): New function, split out from...
28451         (vect_is_simple_reduction): ...here.  Accept reductions that
28452         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28453         (vect_force_simple_reduction): Also store the reduction type in
28454         the assignment's STMT_VINFO_REDUC_TYPE.
28455         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28456         (merge_with_identity): New function.
28457         (vect_expand_fold_left): Likewise.
28458         (vectorize_fold_left_reduction): Likewise.
28459         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28460         scalar phi in place for it.  Check for target support and reject
28461         cases that would reassociate the operation.  Defer the transform
28462         phase to vectorize_fold_left_reduction.
28463         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28464         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28465         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28467 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28469         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28470         call to ifc_temp_var.
28472 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28473             Alan Hayward  <alan.hayward@arm.com>
28474             David Sherwood  <david.sherwood@arm.com>
28476         * target.def (legitimize_address_displacement): Take the original
28477         offset as a poly_int.
28478         * targhooks.h (default_legitimize_address_displacement): Update
28479         accordingly.
28480         * targhooks.c (default_legitimize_address_displacement): Likewise.
28481         * doc/tm.texi: Regenerate.
28482         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28483         as an argument, moving assert of ad->disp == ad->disp_term to...
28484         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28485         Try calling targetm.legitimize_address_displacement before expanding
28486         the address rather than afterwards, and adjust for the new interface.
28487         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28488         Match the new hook interface.  Handle SVE addresses.
28489         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28490         new hook interface.
28492 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28494         * Makefile.in (OBJS): Add early-remat.o.
28495         * target.def (select_early_remat_modes): New hook.
28496         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28497         * doc/tm.texi: Regenerate.
28498         * targhooks.h (default_select_early_remat_modes): Declare.
28499         * targhooks.c (default_select_early_remat_modes): New function.
28500         * timevar.def (TV_EARLY_REMAT): New timevar.
28501         * passes.def (pass_early_remat): New pass.
28502         * tree-pass.h (make_pass_early_remat): Declare.
28503         * early-remat.c: New file.
28504         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28505         function.
28506         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28508 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28509             Alan Hayward  <alan.hayward@arm.com>
28510             David Sherwood  <david.sherwood@arm.com>
28512         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28513         vfm1 with a bound_epilog parameter.
28514         (vect_do_peeling): Update calls accordingly, and move the prologue
28515         call earlier in the function.  Treat the base bound_epilog as 0 for
28516         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28517         this base when peeling for gaps.
28518         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28519         with fully-masked loops.
28520         (vect_estimate_min_profitable_iters): Handle the single peeled
28521         iteration in that case.
28523 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28524             Alan Hayward  <alan.hayward@arm.com>
28525             David Sherwood  <david.sherwood@arm.com>
28527         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28528         single-element interleaving even if the size is not a power of 2.
28529         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28530         accesses for single-element interleaving if the group size is
28531         not a power of 2.
28533 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28534             Alan Hayward  <alan.hayward@arm.com>
28535             David Sherwood  <david.sherwood@arm.com>
28537         * doc/md.texi (fold_extract_last_@var{m}): Document.
28538         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28539         * optabs.def (fold_extract_last_optab): New optab.
28540         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28541         * internal-fn.c (fold_extract_direct): New macro.
28542         (expand_fold_extract_optab_fn): Likewise.
28543         (direct_fold_extract_optab_supported_p): Likewise.
28544         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28545         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28546         EXTRACT_LAST_REDUCTION.
28547         (get_initial_def_for_reduction): Do not create an initial vector
28548         for EXTRACT_LAST_REDUCTION reductions.
28549         (vectorizable_reduction): Leave the scalar phi in place for
28550         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28551         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28552         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28553         transform phase to vectorizable_condition.
28554         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28555         split out from...
28556         (vect_finish_stmt_generation): ...here.
28557         (vect_finish_replace_stmt): New function.
28558         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28559         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28560         pattern.
28561         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28563 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28564             Alan Hayward  <alan.hayward@arm.com>
28565             David Sherwood  <david.sherwood@arm.com>
28567         * doc/md.texi (extract_last_@var{m}): Document.
28568         * optabs.def (extract_last_optab): New optab.
28569         * internal-fn.def (EXTRACT_LAST): New internal function.
28570         * internal-fn.c (cond_unary_direct): New macro.
28571         (expand_cond_unary_optab_fn): Likewise.
28572         (direct_cond_unary_optab_supported_p): Likewise.
28573         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28574         loops using EXTRACT_LAST.
28575         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28576         (extract_last_<mode>): ...this optab.
28577         (vec_extract<mode><Vel>): Update accordingly.
28579 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28580             Alan Hayward  <alan.hayward@arm.com>
28581             David Sherwood  <david.sherwood@arm.com>
28583         * target.def (empty_mask_is_expensive): New hook.
28584         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28585         * doc/tm.texi: Regenerate.
28586         * targhooks.h (default_empty_mask_is_expensive): Declare.
28587         * targhooks.c (default_empty_mask_is_expensive): New function.
28588         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28589         if the target says that empty masks are expensive.
28590         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28591         New function.
28592         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28594 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28595             Alan Hayward  <alan.hayward@arm.com>
28596             David Sherwood  <david.sherwood@arm.com>
28598         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28599         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28600         (vect_use_loop_mask_for_alignment_p): New function.
28601         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28602         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28603         niters_skip argument.  Make sure that the first niters_skip elements
28604         of the first iteration are inactive.
28605         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28606         Update call to vect_set_loop_masks_directly.
28607         (get_misalign_in_elems): New function, split out from...
28608         (vect_gen_prolog_loop_niters): ...here.
28609         (vect_update_init_of_dr): Take a code argument that specifies whether
28610         the adjustment should be added or subtracted.
28611         (vect_update_init_of_drs): Likewise.
28612         (vect_prepare_for_masked_peels): New function.
28613         (vect_do_peeling): Skip prologue peeling if we're using a mask
28614         instead.  Update call to vect_update_inits_of_drs.
28615         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28616         mask_skip_niters.
28617         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28618         alignment.  Do not include the number of peeled iterations in
28619         the minimum threshold in that case.
28620         (vectorizable_induction): Adjust the start value down by
28621         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28622         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28623         Take the number of skipped iterations into account when calculating
28624         the loop bounds.
28625         * tree-vect-stmts.c (vect_gen_while_not): New function.
28627 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28628             Alan Hayward  <alan.hayward@arm.com>
28629             David Sherwood  <david.sherwood@arm.com>
28631         * doc/sourcebuild.texi (vect_fully_masked): Document.
28632         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28633         default value to 0.
28634         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28635         split out from...
28636         (vect_analyze_loop_2): ...here. Don't check the vectorization
28637         factor against the number of loop iterations if the loop is
28638         fully-masked.
28640 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28641             Alan Hayward  <alan.hayward@arm.com>
28642             David Sherwood  <david.sherwood@arm.com>
28644         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28645         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28646         (dump_groups): Update accordingly.
28647         (iv_use::mem_type): New member variable.
28648         (address_p): New function.
28649         (record_use): Add a mem_type argument and initialize the new
28650         mem_type field.
28651         (record_group_use): Add a mem_type argument.  Use address_p.
28652         Remove obsolete null checks of base_object.  Update call to record_use.
28653         (find_interesting_uses_op): Update call to record_group_use.
28654         (find_interesting_uses_cond): Likewise.
28655         (find_interesting_uses_address): Likewise.
28656         (get_mem_type_for_internal_fn): New function.
28657         (find_address_like_use): Likewise.
28658         (find_interesting_uses_stmt): Try find_address_like_use before
28659         calling find_interesting_uses_op.
28660         (addr_offset_valid_p): Use the iv mem_type field as the type
28661         of the addressed memory.
28662         (add_autoinc_candidates): Likewise.
28663         (get_address_cost): Likewise.
28664         (split_small_address_groups_p): Use address_p.
28665         (split_address_groups): Likewise.
28666         (add_iv_candidate_for_use): Likewise.
28667         (autoinc_possible_for_pair): Likewise.
28668         (rewrite_groups): Likewise.
28669         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28670         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28671         (get_alias_ptr_type_for_ptr_address): New function.
28672         (rewrite_use_address): Rewrite address uses in calls that were
28673         identified by find_address_like_use.
28675 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28676             Alan Hayward  <alan.hayward@arm.com>
28677             David Sherwood  <david.sherwood@arm.com>
28679         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28680         TARGET_MEM_REFs.
28681         * gimple-expr.h (is_gimple_addressable: Likewise.
28682         * gimple-expr.c (is_gimple_address): Likewise.
28683         * internal-fn.c (expand_call_mem_ref): New function.
28684         (expand_mask_load_optab_fn): Use it.
28685         (expand_mask_store_optab_fn): Likewise.
28687 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28688             Alan Hayward  <alan.hayward@arm.com>
28689             David Sherwood  <david.sherwood@arm.com>
28691         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28692         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28693         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28694         (cond_umax@var{mode}): Document.
28695         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28696         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28697         (cond_umin_optab, cond_umax_optab): New optabs.
28698         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28699         (COND_IOR, COND_XOR): New internal functions.
28700         * internal-fn.h (get_conditional_internal_fn): Declare.
28701         * internal-fn.c (cond_binary_direct): New macro.
28702         (expand_cond_binary_optab_fn): Likewise.
28703         (direct_cond_binary_optab_supported_p): Likewise.
28704         (get_conditional_internal_fn): New function.
28705         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28706         Cope with reduction statements that are vectorized as calls rather
28707         than assignments.
28708         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28709         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28710         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28711         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28712         (UNSPEC_COND_EOR): New unspecs.
28713         (optab): Add mappings for them.
28714         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28715         (sve_int_op, sve_fp_op): New int attributes.
28717 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28718             Alan Hayward  <alan.hayward@arm.com>
28719             David Sherwood  <david.sherwood@arm.com>
28721         * optabs.def (while_ult_optab): New optab.
28722         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28723         * internal-fn.def (WHILE_ULT): New internal function.
28724         * internal-fn.h (direct_internal_fn_supported_p): New override
28725         that takes two types as argument.
28726         * internal-fn.c (while_direct): New macro.
28727         (expand_while_optab_fn): New function.
28728         (convert_optab_supported_p): Likewise.
28729         (direct_while_optab_supported_p): New macro.
28730         * wide-int.h (wi::udiv_ceil): New function.
28731         * tree-vectorizer.h (rgroup_masks): New structure.
28732         (vec_loop_masks): New typedef.
28733         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28734         and fully_masked_p.
28735         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28736         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28737         (vect_max_vf): New function.
28738         (slpeel_make_loop_iterate_ntimes): Delete.
28739         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28740         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28741         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28742         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28743         internal-fn.h, stor-layout.h and optabs-query.h.
28744         (vect_set_loop_mask): New function.
28745         (add_preheader_seq): Likewise.
28746         (add_header_seq): Likewise.
28747         (interleave_supported_p): Likewise.
28748         (vect_maybe_permute_loop_masks): Likewise.
28749         (vect_set_loop_masks_directly): Likewise.
28750         (vect_set_loop_condition_masked): Likewise.
28751         (vect_set_loop_condition_unmasked): New function, split out from
28752         slpeel_make_loop_iterate_ntimes.
28753         (slpeel_make_loop_iterate_ntimes): Rename to..
28754         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28755         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28756         (vect_do_peeling): Update call accordingly.
28757         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28758         loops.
28759         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28760         mask_compare_type, can_fully_mask_p and fully_masked_p.
28761         (release_vec_loop_masks): New function.
28762         (_loop_vec_info): Use it to free the loop masks.
28763         (can_produce_all_loop_masks_p): New function.
28764         (vect_get_max_nscalars_per_iter): Likewise.
28765         (vect_verify_full_masking): Likewise.
28766         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28767         retries, and free the mask rgroups before retrying.  Check loop-wide
28768         reasons for disallowing fully-masked loops.  Make the final decision
28769         about whether use a fully-masked loop or not.
28770         (vect_estimate_min_profitable_iters): Do not assume that peeling
28771         for the number of iterations will be needed for fully-masked loops.
28772         (vectorizable_reduction): Disable fully-masked loops.
28773         (vectorizable_live_operation): Likewise.
28774         (vect_halve_mask_nunits): New function.
28775         (vect_double_mask_nunits): Likewise.
28776         (vect_record_loop_mask): Likewise.
28777         (vect_get_loop_mask): Likewise.
28778         (vect_transform_loop): Handle the case in which the final loop
28779         iteration might handle a partial vector.  Call vect_set_loop_condition
28780         instead of slpeel_make_loop_iterate_ntimes.
28781         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28782         (check_load_store_masking): New function.
28783         (prepare_load_store_mask): Likewise.
28784         (vectorizable_store): Handle fully-masked loops.
28785         (vectorizable_load): Likewise.
28786         (supportable_widening_operation): Use vect_halve_mask_nunits for
28787         booleans.
28788         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28789         (vect_gen_while): New function.
28790         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28791         (aarch64_uqdec<mode>): New insn.
28793 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28794             Alan Hayward  <alan.hayward@arm.com>
28795             David Sherwood  <david.sherwood@arm.com>
28797         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28798         (reduc_xor_scal_optab): New optabs.
28799         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28800         (reduc_xor_scal_@var{m}): Document.
28801         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28802         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28803         internal functions.
28804         * fold-const-call.c (fold_const_call): Handle them.
28805         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28806         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28807         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28808         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28809         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28810         (UNSPEC_XORV): New unspecs.
28811         (optab): Add entries for them.
28812         (BITWISEV): New int iterator.
28813         (bit_reduc_op): New int attributes.
28815 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28816             Alan Hayward  <alan.hayward@arm.com>
28817             David Sherwood  <david.sherwood@arm.com>
28819         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28820         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28821         * optabs.def (vec_shl_insert_optab): New optab.
28822         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28823         (duplicate_and_interleave): Likewise.
28824         * tree-vect-loop.c: Include internal-fn.h.
28825         (neutral_op_for_slp_reduction): New function, split out from
28826         get_initial_defs_for_reduction.
28827         (get_initial_def_for_reduction): Handle option 2 for variable-length
28828         vectors by loading the neutral value into a vector and then shifting
28829         the initial value into element 0.
28830         (get_initial_defs_for_reduction): Replace the code argument with
28831         the neutral value calculated by neutral_op_for_slp_reduction.
28832         Use gimple_build_vector for constant-length vectors.
28833         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28834         but the first group_size elements have a neutral value.
28835         Use duplicate_and_interleave otherwise.
28836         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28837         Update call to get_initial_defs_for_reduction.  Handle SLP
28838         reductions for variable-length vectors by creating one vector
28839         result for each scalar result, with the elements associated
28840         with other scalar results stubbed out with the neutral value.
28841         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28842         Require IFN_VEC_SHL_INSERT for double reductions on
28843         variable-length vectors, or SLP reductions that have
28844         a neutral value.  Require can_duplicate_and_interleave_p
28845         support for variable-length unchained SLP reductions if there
28846         is no neutral value, such as for MIN/MAX reductions.  Also require
28847         the number of vector elements to be a multiple of the number of
28848         SLP statements when doing variable-length unchained SLP reductions.
28849         Update call to vect_create_epilog_for_reduction.
28850         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28851         and remove initial values.
28852         (duplicate_and_interleave): Make public.
28853         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28854         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28856 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28857             Alan Hayward  <alan.hayward@arm.com>
28858             David Sherwood  <david.sherwood@arm.com>
28860         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28861         (can_duplicate_and_interleave_p): New function.
28862         (vect_get_and_check_slp_defs): Take the vector of statements
28863         rather than just the current one.  Remove excess parentheses.
28864         Restriction rejectinon of vect_constant_def and vect_external_def
28865         for variable-length vectors to boolean types, or types for which
28866         can_duplicate_and_interleave_p is false.
28867         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28868         (duplicate_and_interleave): New function.
28869         (vect_get_constant_vectors): Use gimple_build_vector for
28870         constant-length vectors and suitable variable-length constant
28871         vectors.  Use duplicate_and_interleave for other variable-length
28872         vectors.  Don't defer the update when inserting new statements.
28874 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28875             Alan Hayward  <alan.hayward@arm.com>
28876             David Sherwood  <david.sherwood@arm.com>
28878         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28879         min_profitable_iters doesn't go negative.
28881 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28882             Alan Hayward  <alan.hayward@arm.com>
28883             David Sherwood  <david.sherwood@arm.com>
28885         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28886         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28887         * optabs.def (vec_mask_load_lanes_optab): New optab.
28888         (vec_mask_store_lanes_optab): Likewise.
28889         * internal-fn.def (MASK_LOAD_LANES): New internal function.
28890         (MASK_STORE_LANES): Likewise.
28891         * internal-fn.c (mask_load_lanes_direct): New macro.
28892         (mask_store_lanes_direct): Likewise.
28893         (expand_mask_load_optab_fn): Handle masked operations.
28894         (expand_mask_load_lanes_optab_fn): New macro.
28895         (expand_mask_store_optab_fn): Handle masked operations.
28896         (expand_mask_store_lanes_optab_fn): New macro.
28897         (direct_mask_load_lanes_optab_supported_p): Likewise.
28898         (direct_mask_store_lanes_optab_supported_p): Likewise.
28899         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28900         parameter.
28901         (vect_load_lanes_supported): Likewise.
28902         * tree-vect-data-refs.c (strip_conversion): New function.
28903         (can_group_stmts_p): Likewise.
28904         (vect_analyze_data_ref_accesses): Use it instead of checking
28905         for a pair of assignments.
28906         (vect_store_lanes_supported): Take a masked_p parameter.
28907         (vect_load_lanes_supported): Likewise.
28908         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28909         vect_store_lanes_supported and vect_load_lanes_supported.
28910         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28911         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28912         parameter.  Don't allow gaps for masked accesses.
28913         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
28914         and vect_load_lanes_supported.
28915         (get_load_store_type): Take a masked_p parameter and update
28916         call to get_group_load_store_type.
28917         (vectorizable_store): Update call to get_load_store_type.
28918         Handle IFN_MASK_STORE_LANES.
28919         (vectorizable_load): Update call to get_load_store_type.
28920         Handle IFN_MASK_LOAD_LANES.
28922 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28923             Alan Hayward  <alan.hayward@arm.com>
28924             David Sherwood  <david.sherwood@arm.com>
28926         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28927         modes for SVE.
28928         * config/aarch64/aarch64-protos.h
28929         (aarch64_sve_struct_memory_operand_p): Declare.
28930         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28931         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28932         (VPRED, vpred): Handle SVE structure modes.
28933         * config/aarch64/constraints.md (Utx): New constraint.
28934         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28935         (aarch64_sve_struct_nonimmediate_operand): New predicates.
28936         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28937         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28938         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28939         structure modes.  Split into pieces after RA.
28940         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28941         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28942         New patterns.
28943         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28944         SVE structure modes.
28945         (aarch64_classify_address): Likewise.
28946         (sizetochar): Move earlier in file.
28947         (aarch64_print_operand): Handle SVE register lists.
28948         (aarch64_array_mode): New function.
28949         (aarch64_sve_struct_memory_operand_p): Likewise.
28950         (TARGET_ARRAY_MODE): Redefine.
28952 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28953             Alan Hayward  <alan.hayward@arm.com>
28954             David Sherwood  <david.sherwood@arm.com>
28956         * target.def (array_mode): New target hook.
28957         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28958         * doc/tm.texi: Regenerate.
28959         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28960         * hooks.c (hook_optmode_mode_uhwi_none): New function.
28961         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28962         targetm.array_mode.
28963         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
28964         type sizes.
28966 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28967             Alan Hayward  <alan.hayward@arm.com>
28968             David Sherwood  <david.sherwood@arm.com>
28970         * fold-const.c (fold_binary_loc): Check the argument types
28971         rather than the result type when testing for a vector operation.
28973 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28975         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28976         * doc/tm.texi: Regenerate.
28978 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28979             Alan Hayward  <alan.hayward@arm.com>
28980             David Sherwood  <david.sherwood@arm.com>
28982         * doc/invoke.texi (-msve-vector-bits=): Document new option.
28983         (sve): Document new AArch64 extension.
28984         * doc/md.texi (w): Extend the description of the AArch64
28985         constraint to include SVE vectors.
28986         (Upl, Upa): Document new AArch64 predicate constraints.
28987         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28988         enum.
28989         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28990         (msve-vector-bits=): New option.
28991         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28992         SVE when these are disabled.
28993         (sve): New extension.
28994         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28995         modes.  Adjust their number of units based on aarch64_sve_vg.
28996         (MAX_BITSIZE_MODE_ANY_MODE): Define.
28997         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28998         aarch64_addr_query_type.
28999         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29000         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29001         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29002         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29003         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29004         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29005         (aarch64_simd_imm_zero_p): Delete.
29006         (aarch64_check_zero_based_sve_index_immediate): Declare.
29007         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29008         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29009         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29010         (aarch64_sve_float_mul_immediate_p): Likewise.
29011         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29012         rather than an rtx.
29013         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29014         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29015         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29016         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29017         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29018         (aarch64_regmode_natural_size): Likewise.
29019         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29020         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29021         left one place.
29022         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29023         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29024         for VG and the SVE predicate registers.
29025         (V_ALIASES): Add a "z"-prefixed alias.
29026         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29027         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29028         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29029         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29030         (REG_CLASS_NAMES): Add entries for them.
29031         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29032         and the predicate registers.
29033         (aarch64_sve_vg): Declare.
29034         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29035         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29036         (REGMODE_NATURAL_SIZE): Define.
29037         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29038         SVE macros.
29039         * config/aarch64/aarch64.c: Include cfgrtl.h.
29040         (simd_immediate_info): Add a constructor for series vectors,
29041         and an associated step field.
29042         (aarch64_sve_vg): New variable.
29043         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29044         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29045         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29046         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29047         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29048         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29049         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29050         (aarch64_get_mask_mode): New functions.
29051         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29052         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29053         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29054         predicate modes and predicate registers.  Explicitly restrict
29055         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29056         to store a vector mode if it is recognized by
29057         aarch64_classify_vector_mode.
29058         (aarch64_regmode_natural_size): New function.
29059         (aarch64_hard_regno_caller_save_mode): Return the original mode
29060         for predicates.
29061         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29062         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29063         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29064         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29065         functions.
29066         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29067         does not overlap dest if the function is frame-related.  Handle
29068         SVE constants.
29069         (aarch64_split_add_offset): New function.
29070         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29071         them aarch64_add_offset.
29072         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29073         and update call to aarch64_sub_sp.
29074         (aarch64_add_cfa_expression): New function.
29075         (aarch64_expand_prologue): Pass extra temporary registers to the
29076         functions above.  Handle the case in which we need to emit new
29077         DW_CFA_expressions for registers that were originally saved
29078         relative to the stack pointer, but now have to be expressed
29079         relative to the frame pointer.
29080         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29081         functions above.
29082         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29083         IP0 and IP1 values for SVE frames.
29084         (aarch64_expand_vec_series): New function.
29085         (aarch64_expand_sve_widened_duplicate): Likewise.
29086         (aarch64_expand_sve_const_vector): Likewise.
29087         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29088         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29089         into the register, rather than emitting a SET directly.
29090         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29091         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29092         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29093         (offset_9bit_signed_scaled_p): New functions.
29094         (aarch64_replicate_bitmask_imm): New function.
29095         (aarch64_bitmask_imm): Use it.
29096         (aarch64_cannot_force_const_mem): Reject expressions involving
29097         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29098         (aarch64_classify_index): Handle SVE indices, by requiring
29099         a plain register index with a scale that matches the element size.
29100         (aarch64_classify_address): Handle SVE addresses.  Assert that
29101         the mode of the address is VOIDmode or an integer mode.
29102         Update call to aarch64_classify_symbol.
29103         (aarch64_classify_symbolic_expression): Update call to
29104         aarch64_classify_symbol.
29105         (aarch64_const_vec_all_in_range_p): New function.
29106         (aarch64_print_vector_float_operand): Likewise.
29107         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29108         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29109         and the FP immediates 1.0 and 0.5.
29110         (aarch64_print_address_internal): Handle SVE addresses.
29111         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29112         (aarch64_regno_regclass): Handle predicate registers.
29113         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29114         data modes.
29115         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29116         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29117         (aarch64_convert_sve_vector_bits): New function.
29118         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29119         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29120         rather than an rtx.
29121         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29122         Handle SVE vector and predicate modes.  Accept VL-based constants
29123         that need only one temporary register, and VL offsets that require
29124         no temporary registers.
29125         (aarch64_conditional_register_usage): Mark the predicate registers
29126         as fixed if SVE isn't available.
29127         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29128         Return true for SVE vector and predicate modes.
29129         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29130         rather than an unsigned int.  Handle SVE modes.
29131         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29132         SVE modes.
29133         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29134         if SVE is enabled.
29135         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29136         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29137         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29138         (aarch64_sve_float_mul_immediate_p): New functions.
29139         (aarch64_sve_valid_immediate): New function.
29140         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29141         Explicitly reject structure modes.  Check for INDEX constants.
29142         Handle PTRUE and PFALSE constants.
29143         (aarch64_check_zero_based_sve_index_immediate): New function.
29144         (aarch64_simd_imm_zero_p): Delete.
29145         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29146         vector modes.  Accept constants in the range of CNT[BHWD].
29147         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29148         ask for an Advanced SIMD mode.
29149         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29150         (aarch64_simd_vector_alignment): Handle SVE predicates.
29151         (aarch64_vectorize_preferred_vector_alignment): New function.
29152         (aarch64_simd_vector_alignment_reachable): Use it instead of
29153         the vector size.
29154         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29155         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29156         functions.
29157         (MAX_VECT_LEN): Delete.
29158         (expand_vec_perm_d): Add a vec_flags field.
29159         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29160         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29161         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29162         for SVE modes.
29163         (aarch64_evpc_rev): Rename to...
29164         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29165         (aarch64_evpc_rev_global): New function.
29166         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29167         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29168         MAX_VECT_LEN.
29169         (aarch64_evpc_sve_tbl): New function.
29170         (aarch64_expand_vec_perm_const_1): Update after rename of
29171         aarch64_evpc_rev.  Handle SVE permutes too, trying
29172         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29173         than aarch64_evpc_tbl.
29174         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29175         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29176         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29177         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29178         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29179         (aarch64_expand_sve_vcond): New functions.
29180         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29181         of aarch64_vector_mode_p.
29182         (aarch64_dwarf_poly_indeterminate_value): New function.
29183         (aarch64_compute_pressure_classes): Likewise.
29184         (aarch64_can_change_mode_class): Likewise.
29185         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29186         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29187         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29188         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29189         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29190         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29191         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29192         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29193         constraints.
29194         (Dn, Dl, Dr): Accept const as well as const_vector.
29195         (Dz): Likewise.  Compare against CONST0_RTX.
29196         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29197         of "vector" where appropriate.
29198         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29199         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29200         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29201         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29202         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29203         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29204         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29205         (v_int_equiv): Extend to SVE modes.
29206         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29207         mode attributes.
29208         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29209         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29210         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29211         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29212         (SVE_COND_FP_CMP): New int iterators.
29213         (perm_hilo): Handle the new unpack unspecs.
29214         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29215         attributes.
29216         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29217         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29218         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29219         (aarch64_equality_operator, aarch64_constant_vector_operand)
29220         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29221         (aarch64_sve_nonimmediate_operand): Likewise.
29222         (aarch64_sve_general_operand): Likewise.
29223         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29224         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29225         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29226         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29227         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29228         (aarch64_sve_float_arith_immediate): Likewise.
29229         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29230         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29231         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29232         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29233         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29234         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29235         (aarch64_sve_float_arith_operand): Likewise.
29236         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29237         (aarch64_sve_float_mul_operand): Likewise.
29238         (aarch64_sve_vec_perm_operand): Likewise.
29239         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29240         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29241         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29242         as well as const_vector.
29243         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29244         in file.  Use CONST0_RTX and CONSTM1_RTX.
29245         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29246         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29247         Use aarch64_simd_imm_zero.
29248         * config/aarch64/aarch64-sve.md: New file.
29249         * config/aarch64/aarch64.md: Include it.
29250         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29251         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29252         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29253         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29254         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29255         (sve): New attribute.
29256         (enabled): Disable instructions with the sve attribute unless
29257         TARGET_SVE.
29258         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29259         aarch64_expand_mov_immediate.
29260         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29261         CNT[BHSD] immediates.
29262         (movti): Split CONST_POLY_INT moves into two halves.
29263         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29264         Split additions that need a temporary here if the destination
29265         is the stack pointer.
29266         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29267         (*add<mode>3_poly_1): New instruction.
29268         (set_clobber_cc): New expander.
29270 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29272         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29273         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29274         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29275         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29276         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29277         Change innermode from fixed_mode_size to machine_mode.
29278         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29279         subreg of a variable-length CONST_VECTOR.
29281 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29282             Alan Hayward  <alan.hayward@arm.com>
29283             David Sherwood  <david.sherwood@arm.com>
29285         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29286         (add_offset_to_base): New function, split out from...
29287         (create_mem_ref): ...here.  When handling a scale other than 1,
29288         check first whether the address is valid without the offset.
29289         Add it into the base if so, leaving the index and scale as-is.
29291 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29293         PR c++/83778
29294         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29295         fold_for_warn before checking if arg2 is INTEGER_CST.
29297 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29299         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29300         (store_multiple_operation): Delete.
29301         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29302         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29303         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29304         guarded by TARGET_STRING.
29305         (rs6000_output_load_multiple): Delete.
29306         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29307         OPTION_MASK_STRING / TARGET_STRING handling.
29308         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29309         (const rs6000_opt_masks) <"string">: Change mask to 0.
29310         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29311         (MASK_STRING): Delete.
29312         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29313         parts.  Simplify.
29314         (load_multiple): Delete.
29315         (*ldmsi8): Delete.
29316         (*ldmsi7): Delete.
29317         (*ldmsi6): Delete.
29318         (*ldmsi5): Delete.
29319         (*ldmsi4): Delete.
29320         (*ldmsi3): Delete.
29321         (store_multiple): Delete.
29322         (*stmsi8): Delete.
29323         (*stmsi7): Delete.
29324         (*stmsi6): Delete.
29325         (*stmsi5): Delete.
29326         (*stmsi4): Delete.
29327         (*stmsi3): Delete.
29328         (movmemsi_8reg): Delete.
29329         (corresponding unnamed define_insn): Delete.
29330         (movmemsi_6reg): Delete.
29331         (corresponding unnamed define_insn): Delete.
29332         (movmemsi_4reg): Delete.
29333         (corresponding unnamed define_insn): Delete.
29334         (movmemsi_2reg): Delete.
29335         (corresponding unnamed define_insn): Delete.
29336         (movmemsi_1reg): Delete.
29337         (corresponding unnamed define_insn): Delete.
29338         * config/rs6000/rs6000.opt (mno-string): New.
29339         (mstring): Replace by deprecation warning stub.
29340         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29342 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29344         * regrename.c (regrename_do_replace): If replacing the same
29345         reg multiple times, try to reuse last created gen_raw_REG.
29347         PR debug/81155
29348         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29349         main to workaround a bug in GDB.
29351 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29353         PR target/83737
29354         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29356 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29358         PR rtl-optimization/80481
29359         * ira-color.c (get_cap_member): New function.
29360         (allocnos_conflict_by_live_ranges_p): Use it.
29361         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29362         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29364 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29366         PR target/83628
29367         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29368         (*saddl_se_1): Ditto.
29369         (*ssubsi_1): Ditto.
29370         (*ssubl_se_1): Ditto.
29372 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29374         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29375         rather than wi::to_widest for DR_INITs.
29376         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29377         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29378         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29379         INTEGER_CSTs.
29380         (vect_analyze_group_access_1): Note that here.
29382 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29384         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29385         polynomial type sizes.
29387 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29389         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29390         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29391         (gimple_add_tmp_var): Likewise.
29393 2018-01-12  Martin Liska  <mliska@suse.cz>
29395         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29396         (gimple_alloc_sizes): Likewise.
29397         (dump_gimple_statistics): Use PRIu64 in printf format.
29398         * gimple.h: Change uint64_t to int.
29400 2018-01-12  Martin Liska  <mliska@suse.cz>
29402         * tree-core.h: Use uint64_t instead of int.
29403         * tree.c (tree_node_counts): Likewise.
29404         (tree_node_sizes): Likewise.
29405         (dump_tree_statistics): Use PRIu64 in printf format.
29407 2018-01-12  Martin Liska  <mliska@suse.cz>
29409         * Makefile.in: As qsort_chk is implemented in vec.c, add
29410         vec.o to linkage of gencfn-macros.
29411         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29412         passing the info to record_node_allocation_statistics.
29413         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29414         and pass the info.
29415         * ggc-common.c (struct ggc_usage): Add operator== and use
29416         it in operator< and compare function.
29417         * mem-stats.h (struct mem_usage): Likewise.
29418         * vec.c (struct vec_usage): Remove operator< and compare
29419         function. Can be simply inherited.
29421 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29423         PR target/81616
29424         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29425         * tree-ssa-math-opts.c: Include domwalk.h.
29426         (convert_mult_to_fma_1): New function.
29427         (fma_transformation_info): New type.
29428         (fma_deferring_state): Likewise.
29429         (cancel_fma_deferring): New function.
29430         (result_of_phi): Likewise.
29431         (last_fma_candidate_feeds_initial_phi): Likewise.
29432         (convert_mult_to_fma): Added deferring logic, split actual
29433         transformation to convert_mult_to_fma_1.
29434         (math_opts_dom_walker): New type.
29435         (math_opts_dom_walker::after_dom_children): New method, body moved
29436         here from pass_optimize_widening_mul::execute, added deferring logic
29437         bits.
29438         (pass_optimize_widening_mul::execute): Moved most of code to
29439         math_opts_dom_walker::after_dom_children.
29440         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29441         * config/i386/i386.c (ix86_option_override_internal): Added
29442         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29444 2018-01-12  Richard Biener  <rguenther@suse.de>
29446         PR debug/83157
29447         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29448         inline instance vars.
29450 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29452         PR target/81819
29453         * config/rx/rx.c (rx_is_restricted_memory_address):
29454         Handle SUBREG case.
29456 2018-01-12  Richard Biener  <rguenther@suse.de>
29458         PR tree-optimization/80846
29459         * target.def (split_reduction): New target hook.
29460         * targhooks.c (default_split_reduction): New function.
29461         * targhooks.h (default_split_reduction): Declare.
29462         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29463         target requests first reduce vectors by combining low and high
29464         parts.
29465         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29466         (get_vectype_for_scalar_type_and_size): Export.
29467         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29468         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29469         * doc/tm.texi: Regenerate.
29470         * config/i386/i386.c (ix86_split_reduction): Implement
29471         TARGET_VECTORIZE_SPLIT_REDUCTION.
29473 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29475         PR target/83368
29476         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29477         in PIC mode except for TARGET_VXWORKS_RTP.
29478         * config/sparc/sparc.c: Include cfgrtl.h.
29479         (TARGET_INIT_PIC_REG): Define.
29480         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29481         (sparc_pic_register_p): New predicate.
29482         (sparc_legitimate_address_p): Use it.
29483         (sparc_legitimize_pic_address): Likewise.
29484         (sparc_delegitimize_address): Likewise.
29485         (sparc_mode_dependent_address_p): Likewise.
29486         (gen_load_pcrel_sym): Remove 4th parameter.
29487         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29488         (sparc_expand_prologue): Do not call load_got_register here.
29489         (sparc_flat_expand_prologue): Likewise.
29490         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29491         (sparc_use_pseudo_pic_reg): New function.
29492         (sparc_init_pic_reg): Likewise.
29493         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29494         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29496 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29498         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29499         Add item for branch_cost.
29501 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29503         PR rtl-optimization/83565
29504         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29505         not extend the result to a larger mode for rotate operations.
29506         (num_sign_bit_copies1): Likewise.
29508 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29510         PR target/40411
29511         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29512         -symbolic.
29513         Use values-Xc.o for -pedantic.
29514         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29516 2018-01-12  Martin Liska  <mliska@suse.cz>
29518         PR ipa/83054
29519         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29520         New function.
29521         (possible_polymorphic_call_targets): Use it.
29522         (ipa_devirt): Likewise.
29524 2018-01-12  Martin Liska  <mliska@suse.cz>
29526         * profile-count.h (enum profile_quality): Use 0 as invalid
29527         enum value of profile_quality.
29529 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29531         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29532         -mext-string options.
29534 2018-01-12  Richard Biener  <rguenther@suse.de>
29536         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29537         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29538         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29539         Likewise.
29540         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29542 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29544         * configure.ac (--with-long-double-format): Add support for the
29545         configuration option to change the default long double format on
29546         PowerPC systems.
29547         * config.gcc (powerpc*-linux*-*): Likewise.
29548         * configure: Regenerate.
29549         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29550         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29551         used without modification.
29553 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29555         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29556         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29557         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29558         MISC_BUILTIN_SPEC_BARRIER.
29559         (rs6000_init_builtins): Likewise.
29560         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29561         enum value.
29562         (speculation_barrier): New define_insn.
29563         * doc/extend.texi: Document __builtin_speculation_barrier.
29565 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29567         PR target/83203
29568         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29569         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29570         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29571         iterators.
29572         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29573         integral modes instead of "ss" and "sd".
29574         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29575         vectors with 32-bit and 64-bit elements.
29576         (vecdupssescalarmodesuffix): New mode attribute.
29577         (vec_dup<mode>): Use it.
29579 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29581         PR target/83330
29582         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29583         frame if argument is passed on stack.
29585 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29587         PR target/82682
29588         * ree.c (combine_reaching_defs): Optimize also
29589         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29590         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29592 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29594         PR middle-end/83189
29595         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29597 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29599         PR middle-end/83718
29600         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29601         after they are computed.
29603 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29605         PR tree-optimization/83695
29606         * gimple-loop-linterchange.cc
29607         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29608         reset cached scev information after interchange.
29609         (pass_linterchange::execute): Remove call to scev_reset_htab.
29611 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29613         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29614         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29615         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29616         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29617         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29618         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29619         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29620         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29621         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29622         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29623         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29624         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29625         (V_lane_reg): Likewise.
29626         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29627         New define_expand.
29628         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29629         (vfmal_lane_low<mode>_intrinsic,
29630         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29631         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29632         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29633         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29634         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29635         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29637 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29639         * config/arm/arm-cpus.in (fp16fml): New feature.
29640         (ALL_SIMD): Add fp16fml.
29641         (armv8.2-a): Add fp16fml as an option.
29642         (armv8.3-a): Likewise.
29643         (armv8.4-a): Add fp16fml as part of fp16.
29644         * config/arm/arm.h (TARGET_FP16FML): Define.
29645         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29646         when appropriate.
29647         * config/arm/arm-modes.def (V2HF): Define.
29648         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29649         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29650         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29651         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29652         vfmsl_low, vfmsl_high): New set of builtins.
29653         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29654         (vfml_op): New code attribute.
29655         (VFMLHALVES): New int iterator.
29656         (VFML, VFMLSEL): New mode attributes.
29657         (V_reg): Define mapping for V2HF.
29658         (V_hi, V_lo): New mode attributes.
29659         (VF_constraint): Likewise.
29660         (vfml_half, vfml_half_selector): New int attributes.
29661         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29662         define_expand.
29663         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29664         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29665         New define_insn.
29666         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29667         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29668         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29669         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29670         documentation.
29671         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29672         Document new effective target and option set.
29674 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29676         * config/arm/arm-cpus.in (armv8_4): New feature.
29677         (ARMv8_4a): New fgroup.
29678         (armv8.4-a): New arch.
29679         * config/arm/arm-tables.opt: Regenerate.
29680         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29681         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29682         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29683         Add matching rules for -march=armv8.4-a and extensions.
29684         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29686 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29688         PR target/81821
29689         * config/rx/rx.md (BW): New mode attribute.
29690         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29692 2018-01-11  Richard Biener  <rguenther@suse.de>
29694         PR tree-optimization/83435
29695         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29696         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29697         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29699 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29700             Alan Hayward  <alan.hayward@arm.com>
29701             David Sherwood  <david.sherwood@arm.com>
29703         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29704         field.
29705         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29706         (aarch64_print_address_internal): Use it to check for a zero offset.
29708 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29709             Alan Hayward  <alan.hayward@arm.com>
29710             David Sherwood  <david.sherwood@arm.com>
29712         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29713         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29714         Return a poly_int64 rather than a HOST_WIDE_INT.
29715         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29716         rather than a HOST_WIDE_INT.
29717         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29718         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29719         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29720         final_offset from HOST_WIDE_INT to poly_int64.
29721         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29722         to_constant when getting the number of units in an Advanced SIMD
29723         mode.
29724         (aarch64_builtin_vectorized_function): Check for a constant number
29725         of units.
29726         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29727         GET_MODE_SIZE.
29728         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29729         attribute instead of GET_MODE_NUNITS.
29730         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29731         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29732         GET_MODE_SIZE for fixed-size registers.
29733         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29734         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29735         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29736         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29737         (aarch64_print_operand, aarch64_print_address_internal)
29738         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29739         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29740         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29741         Handle polynomial GET_MODE_SIZE.
29742         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29743         wider than SImode without modification.
29744         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29745         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29746         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29747         passing and returning SVE modes.
29748         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29749         rather than GEN_INT.
29750         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29751         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29752         (aarch64_allocate_and_probe_stack_space): Likewise.
29753         (aarch64_layout_frame): Cope with polynomial offsets.
29754         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29755         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29756         polynomial offsets.
29757         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29758         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29759         poly_int64 rather than a HOST_WIDE_INT.
29760         (aarch64_get_separate_components, aarch64_process_components)
29761         (aarch64_expand_prologue, aarch64_expand_epilogue)
29762         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29763         (aarch64_anchor_offset): New function, split out from...
29764         (aarch64_legitimize_address): ...here.
29765         (aarch64_builtin_vectorization_cost): Handle polynomial
29766         TYPE_VECTOR_SUBPARTS.
29767         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29768         GET_MODE_NUNITS.
29769         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29770         number of elements from the PARALLEL rather than the mode.
29771         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29772         rather than GET_MODE_BITSIZE.
29773         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29774         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29775         (aarch64_expand_vec_perm_const_1): Handle polynomial
29776         d->perm.length () and d->perm elements.
29777         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29778         Apply to_constant to d->perm elements.
29779         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29780         polynomial CONST_VECTOR_NUNITS.
29781         (aarch64_move_pointer): Take amount as a poly_int64 rather
29782         than an int.
29783         (aarch64_progress_pointer): Avoid temporary variable.
29784         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29785         the mode attribute instead of GET_MODE.
29787 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29788             Alan Hayward  <alan.hayward@arm.com>
29789             David Sherwood  <david.sherwood@arm.com>
29791         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29792         x exists before using it.
29793         (aarch64_add_constant_internal): Rename to...
29794         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29795         src and dest rtxes.  Handle the case in which they're different,
29796         including when the offset is zero.  Replace scratchreg with an rtx.
29797         Use 2 additions if there is no spare register into which we can
29798         move a 16-bit constant.
29799         (aarch64_add_constant): Delete.
29800         (aarch64_add_offset): Replace reg with separate src and dest
29801         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29802         Use aarch64_add_offset_1.
29803         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29804         an rtx rather than an int.  Take the delta as a poly_int64
29805         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29806         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29807         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29808         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29809         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29810         and aarch64_add_sp.
29811         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29812         aarch64_add_constant.
29814 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29816         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29817         Use scalar_float_mode.
29819 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29821         * config/aarch64/aarch64-simd.md
29822         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29823         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29824         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29825         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29826         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29827         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29828         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29829         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29830         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29831         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29833 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29835         PR target/83514
29836         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29837         targ_options->x_arm_arch_string is non NULL.
29839 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29841         * config/aarch64/aarch64.h
29842         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29844 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29846         PR target/82096
29847         * expmed.c (emit_store_flag_force): Swap if const op0
29848         and change VOIDmode to mode of op0.
29850 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29852         PR rtl-optimization/83761
29853         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29854         than bytes to mode_for_size.
29856 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29858         PR middle-end/83189
29859         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29860         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29861         profile.
29863 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29865         PR middle-end/83575
29866         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29867         when in layout mode.
29868         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29869         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29870         partition fixup.
29872 2018-01-10  Michael Collison  <michael.collison@arm.com>
29874         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29875         * config/aarch64/aarch64-option-extension.def: Add
29876         AARCH64_OPT_EXTENSION of 'fp16fml'.
29877         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29878         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29879         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29880         * config/aarch64/constraints.md (Ui7): New constraint.
29881         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29882         (VFMLA_SEL_W): Ditto.
29883         (f16quad): Ditto.
29884         (f16mac1): Ditto.
29885         (VFMLA16_LOW): New int iterator.
29886         (VFMLA16_HIGH): Ditto.
29887         (UNSPEC_FMLAL): New unspec.
29888         (UNSPEC_FMLSL): Ditto.
29889         (UNSPEC_FMLAL2): Ditto.
29890         (UNSPEC_FMLSL2): Ditto.
29891         (f16mac): New code attribute.
29892         * config/aarch64/aarch64-simd-builtins.def
29893         (aarch64_fmlal_lowv2sf): Ditto.
29894         (aarch64_fmlsl_lowv2sf): Ditto.
29895         (aarch64_fmlalq_lowv4sf): Ditto.
29896         (aarch64_fmlslq_lowv4sf): Ditto.
29897         (aarch64_fmlal_highv2sf): Ditto.
29898         (aarch64_fmlsl_highv2sf): Ditto.
29899         (aarch64_fmlalq_highv4sf): Ditto.
29900         (aarch64_fmlslq_highv4sf): Ditto.
29901         (aarch64_fmlal_lane_lowv2sf): Ditto.
29902         (aarch64_fmlsl_lane_lowv2sf): Ditto.
29903         (aarch64_fmlal_laneq_lowv2sf): Ditto.
29904         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29905         (aarch64_fmlalq_lane_lowv4sf): Ditto.
29906         (aarch64_fmlsl_lane_lowv4sf): Ditto.
29907         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29908         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29909         (aarch64_fmlal_lane_highv2sf): Ditto.
29910         (aarch64_fmlsl_lane_highv2sf): Ditto.
29911         (aarch64_fmlal_laneq_highv2sf): Ditto.
29912         (aarch64_fmlsl_laneq_highv2sf): Ditto.
29913         (aarch64_fmlalq_lane_highv4sf): Ditto.
29914         (aarch64_fmlsl_lane_highv4sf): Ditto.
29915         (aarch64_fmlalq_laneq_highv4sf): Ditto.
29916         (aarch64_fmlsl_laneq_highv4sf): Ditto.
29917         * config/aarch64/aarch64-simd.md:
29918         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29919         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29920         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29921         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29922         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29923         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29924         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29925         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29926         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29927         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29928         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29929         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29930         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29931         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29932         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29933         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29934         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29935         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29936         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29937         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29938         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29939         (vfmlsl_low_u32): Ditto.
29940         (vfmlalq_low_u32): Ditto.
29941         (vfmlslq_low_u32): Ditto.
29942         (vfmlal_high_u32): Ditto.
29943         (vfmlsl_high_u32): Ditto.
29944         (vfmlalq_high_u32): Ditto.
29945         (vfmlslq_high_u32): Ditto.
29946         (vfmlal_lane_low_u32): Ditto.
29947         (vfmlsl_lane_low_u32): Ditto.
29948         (vfmlal_laneq_low_u32): Ditto.
29949         (vfmlsl_laneq_low_u32): Ditto.
29950         (vfmlalq_lane_low_u32): Ditto.
29951         (vfmlslq_lane_low_u32): Ditto.
29952         (vfmlalq_laneq_low_u32): Ditto.
29953         (vfmlslq_laneq_low_u32): Ditto.
29954         (vfmlal_lane_high_u32): Ditto.
29955         (vfmlsl_lane_high_u32): Ditto.
29956         (vfmlal_laneq_high_u32): Ditto.
29957         (vfmlsl_laneq_high_u32): Ditto.
29958         (vfmlalq_lane_high_u32): Ditto.
29959         (vfmlslq_lane_high_u32): Ditto.
29960         (vfmlalq_laneq_high_u32): Ditto.
29961         (vfmlslq_laneq_high_u32): Ditto.
29962         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29963         (AARCH64_FL_FOR_ARCH8_4): New.
29964         (AARCH64_ISA_F16FML): New ISA flag.
29965         (TARGET_F16FML): New feature flag for fp16fml.
29966         (doc/invoke.texi): Document new fp16fml option.
29968 2018-01-10  Michael Collison  <michael.collison@arm.com>
29970         * config/aarch64/aarch64-builtins.c:
29971         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29972         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29973         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29974         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29975         (AARCH64_ISA_SHA3): New ISA flag.
29976         (TARGET_SHA3): New feature flag for sha3.
29977         * config/aarch64/iterators.md (sha512_op): New int attribute.
29978         (CRYPTO_SHA512): New int iterator.
29979         (UNSPEC_SHA512H): New unspec.
29980         (UNSPEC_SHA512H2): Ditto.
29981         (UNSPEC_SHA512SU0): Ditto.
29982         (UNSPEC_SHA512SU1): Ditto.
29983         * config/aarch64/aarch64-simd-builtins.def
29984         (aarch64_crypto_sha512hqv2di): New builtin.
29985         (aarch64_crypto_sha512h2qv2di): Ditto.
29986         (aarch64_crypto_sha512su0qv2di): Ditto.
29987         (aarch64_crypto_sha512su1qv2di): Ditto.
29988         (aarch64_eor3qv8hi): Ditto.
29989         (aarch64_rax1qv2di): Ditto.
29990         (aarch64_xarqv2di): Ditto.
29991         (aarch64_bcaxqv8hi): Ditto.
29992         * config/aarch64/aarch64-simd.md:
29993         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29994         (aarch64_crypto_sha512su0qv2di): Ditto.
29995         (aarch64_crypto_sha512su1qv2di): Ditto.
29996         (aarch64_eor3qv8hi): Ditto.
29997         (aarch64_rax1qv2di): Ditto.
29998         (aarch64_xarqv2di): Ditto.
29999         (aarch64_bcaxqv8hi): Ditto.
30000         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30001         (vsha512h2q_u64): Ditto.
30002         (vsha512su0q_u64): Ditto.
30003         (vsha512su1q_u64): Ditto.
30004         (veor3q_u16): Ditto.
30005         (vrax1q_u64): Ditto.
30006         (vxarq_u64): Ditto.
30007         (vbcaxq_u16): Ditto.
30008         * config/arm/types.md (crypto_sha512): New type attribute.
30009         (crypto_sha3): Ditto.
30010         (doc/invoke.texi): Document new sha3 option.
30012 2018-01-10  Michael Collison  <michael.collison@arm.com>
30014         * config/aarch64/aarch64-builtins.c:
30015         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30016         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30017         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30018         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30019         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30020         (AARCH64_ISA_SM4): New ISA flag.
30021         (TARGET_SM4): New feature flag for sm4.
30022         * config/aarch64/aarch64-simd-builtins.def
30023         (aarch64_sm3ss1qv4si): Ditto.
30024         (aarch64_sm3tt1aq4si): Ditto.
30025         (aarch64_sm3tt1bq4si): Ditto.
30026         (aarch64_sm3tt2aq4si): Ditto.
30027         (aarch64_sm3tt2bq4si): Ditto.
30028         (aarch64_sm3partw1qv4si): Ditto.
30029         (aarch64_sm3partw2qv4si): Ditto.
30030         (aarch64_sm4eqv4si): Ditto.
30031         (aarch64_sm4ekeyqv4si): Ditto.
30032         * config/aarch64/aarch64-simd.md:
30033         (aarch64_sm3ss1qv4si): Ditto.
30034         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30035         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30036         (aarch64_sm4eqv4si): Ditto.
30037         (aarch64_sm4ekeyqv4si): Ditto.
30038         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30039         (sm3part_op): Ditto.
30040         (CRYPTO_SM3TT): Ditto.
30041         (CRYPTO_SM3PART): Ditto.
30042         (UNSPEC_SM3SS1): New unspec.
30043         (UNSPEC_SM3TT1A): Ditto.
30044         (UNSPEC_SM3TT1B): Ditto.
30045         (UNSPEC_SM3TT2A): Ditto.
30046         (UNSPEC_SM3TT2B): Ditto.
30047         (UNSPEC_SM3PARTW1): Ditto.
30048         (UNSPEC_SM3PARTW2): Ditto.
30049         (UNSPEC_SM4E): Ditto.
30050         (UNSPEC_SM4EKEY): Ditto.
30051         * config/aarch64/constraints.md (Ui2): New constraint.
30052         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30053         * config/arm/types.md (crypto_sm3): New type attribute.
30054         (crypto_sm4): Ditto.
30055         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30056         (vsm3tt1aq_u32): Ditto.
30057         (vsm3tt1bq_u32): Ditto.
30058         (vsm3tt2aq_u32): Ditto.
30059         (vsm3tt2bq_u32): Ditto.
30060         (vsm3partw1q_u32): Ditto.
30061         (vsm3partw2q_u32): Ditto.
30062         (vsm4eq_u32): Ditto.
30063         (vsm4ekeyq_u32): Ditto.
30064         (doc/invoke.texi): Document new sm4 option.
30066 2018-01-10  Michael Collison  <michael.collison@arm.com>
30068         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30069         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30070         (AARCH64_FL_FOR_ARCH8_4): New.
30071         (AARCH64_FL_V8_4): New flag.
30072         (doc/invoke.texi): Document new armv8.4-a option.
30074 2018-01-10  Michael Collison  <michael.collison@arm.com>
30076         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30077         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30078         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30079         * config/aarch64/aarch64-option-extension.def: Add
30080         AARCH64_OPT_EXTENSION of 'sha2'.
30081         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30082         (crypto): Disable sha2 and aes if crypto disabled.
30083         (crypto): Enable aes and sha2 if enabled.
30084         (simd): Disable sha2 and aes if simd disabled.
30085         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30086         New flags.
30087         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30088         (TARGET_SHA2): New feature flag for sha2.
30089         (TARGET_AES): New feature flag for aes.
30090         * config/aarch64/aarch64-simd.md:
30091         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30092         conditional on TARGET_AES.
30093         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30094         (aarch64_crypto_sha1hsi): Make pattern conditional
30095         on TARGET_SHA2.
30096         (aarch64_crypto_sha1hv4si): Ditto.
30097         (aarch64_be_crypto_sha1hv4si): Ditto.
30098         (aarch64_crypto_sha1su1v4si): Ditto.
30099         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30100         (aarch64_crypto_sha1su0v4si): Ditto.
30101         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30102         (aarch64_crypto_sha256su0v4si): Ditto.
30103         (aarch64_crypto_sha256su1v4si): Ditto.
30104         (doc/invoke.texi): Document new aes and sha2 options.
30106 2018-01-10  Martin Sebor  <msebor@redhat.com>
30108         PR tree-optimization/83781
30109         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30110         as string arrays.
30112 2018-01-11  Martin Sebor  <msebor@gmail.com>
30113             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30115         PR tree-optimization/83501
30116         PR tree-optimization/81703
30118         * tree-ssa-strlen.c (get_string_cst): Rename...
30119         (get_string_len): ...to this.  Handle global constants.
30120         (handle_char_store): Adjust.
30122 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30123             Jim Wilson  <jimw@sifive.com>
30125         * config/riscv/riscv-protos.h (riscv_output_return): New.
30126         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30127         (riscv_attribute_table, riscv_output_return),
30128         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30129         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30130         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30131         function.
30132         (riscv_expand_prologue): Add early return for naked function.
30133         (riscv_expand_epilogue): Likewise.
30134         (riscv_function_ok_for_sibcall): Return false for naked function.
30135         (riscv_set_current_function): New.
30136         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30137         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30138         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30139         * doc/extend.texi (RISC-V Function Attributes): New.
30141 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30143         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30144         check for 128-bit long double before checking TCmode.
30145         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30146         128-bit long doubles before checking TFmode or TCmode.
30147         (FLOAT128_IBM_P): Likewise.
30149 2018-01-10  Martin Sebor  <msebor@redhat.com>
30151         PR tree-optimization/83671
30152         * builtins.c (c_strlen): Unconditionally return zero for the empty
30153         string.
30154         Use -Warray-bounds for warnings.
30155         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30156         for non-constant array indices with COMPONENT_REF, arrays of
30157         arrays, and pointers to arrays.
30158         (gimple_fold_builtin_strlen): Determine and set length range for
30159         non-constant character arrays.
30161 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30163         PR middle-end/81897
30164         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30165         empty blocks.
30167 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30169         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30171 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30173         PR target/83399
30174         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30175         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30176         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30177         indexed_or_indirect_operand predicate.
30178         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30179         (*vsx_le_perm_load_v8hi): Likewise.
30180         (*vsx_le_perm_load_v16qi): Likewise.
30181         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30182         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30183         (*vsx_le_perm_store_v8hi): Likewise.
30184         (*vsx_le_perm_store_v16qi): Likewise.
30185         (eight unnamed splitters): Likewise.
30187 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30189         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30190         * config/rs6000/emmintrin.h: Likewise.
30191         * config/rs6000/mmintrin.h: Likewise.
30192         * config/rs6000/xmmintrin.h: Likewise.
30194 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30196         PR c++/43486
30197         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30198         "public_flag".
30199         * tree.c (tree_nop_conversion): Return true for location wrapper
30200         nodes.
30201         (maybe_wrap_with_location): New function.
30202         (selftest::check_strip_nops): New function.
30203         (selftest::test_location_wrappers): New function.
30204         (selftest::tree_c_tests): Call it.
30205         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30206         (maybe_wrap_with_location): New decl.
30207         (EXPR_LOCATION_WRAPPER_P): New macro.
30208         (location_wrapper_p): New inline function.
30209         (tree_strip_any_location_wrapper): New inline function.
30211 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30213         PR target/83735
30214         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30215         stack_realign_offset for the largest alignment of stack slot
30216         actually used.
30217         (ix86_find_max_used_stack_alignment): New function.
30218         (ix86_finalize_stack_frame_flags): Use it.  Set
30219         max_used_stack_alignment if we don't realign stack.
30220         * config/i386/i386.h (machine_function): Add
30221         max_used_stack_alignment.
30223 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30225         * config/arm/arm.opt (-mbranch-cost): New option.
30226         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30227         account.
30229 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30231         PR target/83629
30232         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30233         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30235 2018-01-10  Richard Biener  <rguenther@suse.de>
30237         PR debug/83765
30238         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30239         early out so it also covers the case where we have a non-NULL
30240         origin.
30242 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30244         PR tree-optimization/83753
30245         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30246         for non-strided grouped accesses if the number of elements is 1.
30248 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30250         PR target/81616
30251         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30252         * i386.h (TARGET_USE_GATHER): Define.
30253         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30255 2018-01-10  Martin Liska  <mliska@suse.cz>
30257         PR bootstrap/82831
30258         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30259         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30260         partitioning.
30261         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30262         CLEANUP_NO_PARTITIONING is not set.
30264 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30266         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30267         for vectors, as a partial revert of r254296.
30268         * rtl.h (const_vec_p): Delete.
30269         (const_vec_duplicate_p): Don't test for vector CONSTs.
30270         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30271         * expmed.c (make_tree): Likewise.
30273         Revert:
30274         * common.md (E, F): Use CONSTANT_P instead of checking for
30275         CONST_VECTOR.
30276         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30277         checking for CONST_VECTOR.
30279 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30281         PR middle-end/83575
30282         * predict.c (force_edge_cold): Handle in more sane way edges
30283         with no prediction.
30285 2018-01-09  Carl Love  <cel@us.ibm.com>
30287         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30288         V4SI, V4SF types.
30289         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30290         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30291         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30292         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30293         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30294         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30295         * config/rs6000/rs6000-protos.h: Add extern defition for
30296         rs6000_generate_float2_double_code.
30297         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30298         function.
30299         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30300         (float2_v2df): Add define_expand.
30302 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30304         PR target/83628
30305         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30306         op_mode in the force_to_mode call.
30308 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30310         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30311         instead of checking each element individually.
30312         (aarch64_evpc_uzp): Likewise.
30313         (aarch64_evpc_zip): Likewise.
30314         (aarch64_evpc_ext): Likewise.
30315         (aarch64_evpc_rev): Likewise.
30316         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30317         instead of checking each element individually.  Return true without
30318         generating rtl if
30319         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30320         whether all selected elements come from the same input, instead of
30321         checking each element individually.  Remove calls to gen_rtx_REG,
30322         start_sequence and end_sequence and instead assert that no rtl is
30323         generated.
30325 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30327         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30328         order of HIGH and CONST checks.
30330 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30332         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30333         if the destination isn't an SSA_NAME.
30335 2018-01-09  Richard Biener  <rguenther@suse.de>
30337         PR tree-optimization/83668
30338         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30339         move prologue...
30340         (canonicalize_loop_form): ... here, renamed from ...
30341         (canonicalize_loop_closed_ssa_form): ... this and amended to
30342         swap successor edges for loop exit blocks to make us use
30343         the RPO order we need for initial schedule generation.
30345 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30347         PR tree-optimization/64811
30348         * match.pd: When optimizing comparisons with Inf, avoid
30349         introducing or losing exceptions from comparisons with NaN.
30351 2018-01-09  Martin Liska  <mliska@suse.cz>
30353         PR sanitizer/82517
30354         * asan.c (shadow_mem_size): Add gcc_assert.
30356 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30358         Don't save registers in main().
30360         PR target/83738
30361         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30362         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30363         * config/avr/avr.c (avr_set_current_function): Don't error if
30364         naked, OS_task or OS_main are specified at the same time.
30365         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30366         OS_main.
30367         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30368         attribute.
30369         * common/config/avr/avr-common.c (avr_option_optimization_table):
30370         Switch on -mmain-is-OS_task for optimizing compilations.
30372 2018-01-09  Richard Biener  <rguenther@suse.de>
30374         PR tree-optimization/83572
30375         * graphite.c: Include cfganal.h.
30376         (graphite_transform_loops): Connect infinite loops to exit
30377         and remove fake edges at the end.
30379 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30381         * ipa-inline.c (edge_badness): Revert accidental checkin.
30383 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30385         PR ipa/80763
30386         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30387         symbols; not inline clones.
30389 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30391         PR target/83507
30392         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30393         hard registers.  Formatting fixes.
30395         PR preprocessor/83722
30396         * gcc.c (try_generate_repro): Pass
30397         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30398         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30399         do_report_bug.
30401 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30402             Kito Cheng  <kito.cheng@gmail.com>
30404         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30405         (riscv_leaf_function_p): Delete.
30406         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30408 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30410         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30411         function.
30412         (do_ifelse): New function.
30413         (do_isel): New function.
30414         (do_sub3): New function.
30415         (do_add3): New function.
30416         (do_load_mask_compare): New function.
30417         (do_overlap_load_compare): New function.
30418         (expand_compare_loop): New function.
30419         (expand_block_compare): Call expand_compare_loop() when appropriate.
30420         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30421         option description.
30422         (-mblock-compare-inline-loop-limit): New option.
30424 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30426         PR target/83677
30427         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30428         Reverse order of second and third operands in first alternative.
30429         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30430         of first and second elements in UNSPEC_VPERMR vector.
30431         (altivec_expand_vec_perm_le): Likewise.
30433 2018-01-08  Jeff Law  <law@redhat.com>
30435         PR rtl-optimizatin/81308
30436         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30437         (process_switch): If group_case_labels makes a change, then set
30438         cfg_altered.
30439         (pass_convert_switch::execute): If a switch is converted, then
30440         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30442         PR rtl-optimization/81308
30443         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30444         splitting insns.
30446 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30448         PR target/83663 - Revert r255946
30449         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30450         generation for cases where splatting a value is not useful.
30451         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30452         across a vec_duplicate and a paradoxical subreg forming a vector
30453         mode to a vec_concat.
30455 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30457         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30458         -march=armv8.3-a variants.
30459         * config/arm/t-multilib: Likewise.
30460         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30462 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30464         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30465         to generate rtl.
30466         (cceq_ior_compare_complement): Give it a name so I can use it, and
30467         change boolean_or_operator predicate to boolean_operator so it can
30468         be used to generate a crand.
30469         (eqne): New code iterator.
30470         (bd/bd_neg): New code_attrs.
30471         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30472         a single define_insn.
30473         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30474         decrement (bdnzt/bdnzf/bdzt/bdzf).
30475         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30476         with the new names of the branch decrement patterns, and added the
30477         names of the branch decrement conditional patterns.
30479 2018-01-08  Richard Biener  <rguenther@suse.de>
30481         PR tree-optimization/83563
30482         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30483         cache.
30485 2018-01-08  Richard Biener  <rguenther@suse.de>
30487         PR middle-end/83713
30488         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30490 2018-01-08  Richard Biener  <rguenther@suse.de>
30492         PR tree-optimization/83685
30493         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30494         references to abnormals.
30496 2018-01-08  Richard Biener  <rguenther@suse.de>
30498         PR lto/83719
30499         * dwarf2out.c (output_indirect_strings): Handle empty
30500         skeleton_debug_str_hash.
30501         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30503 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30505         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30506         (emit_store_direct): Likewise.
30507         (arc_trampoline_adjust_address): Likewise.
30508         (arc_asm_trampoline_template): New function.
30509         (arc_initialize_trampoline): Use asm_trampoline_template.
30510         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30511         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30512         * config/arc/arc.md (flush_icache): Delete pattern.
30514 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30516         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30517         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30518         munaligned-access.
30520 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30522         PR target/83681
30523         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30524         by not USED_FOR_TARGET.
30525         (make_pass_resolve_sw_modes): Likewise.
30527 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30529         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30530         USED_FOR_TARGET.
30532 2018-01-08  Richard Biener  <rguenther@suse.de>
30534         PR middle-end/83580
30535         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30537 2018-01-08  Richard Biener  <rguenther@suse.de>
30539         PR middle-end/83517
30540         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30542 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30544         PR middle-end/81897
30545         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30546         basic blocks with a small number of successors.
30547         (convert_control_dep_chain_into_preds): Improve handling of
30548         forwarder blocks.
30549         (dump_predicates): Split apart into...
30550         (dump_pred_chain): ...here...
30551         (dump_pred_info): ...and here.
30552         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30553         (can_chain_union_be_invalidated_p): Improve check for invalidation
30554         of paths.
30555         (uninit_uses_cannot_happen): Avoid unnecessary if
30556         convert_control_dep_chain_into_preds yielded nothing.
30558 2018-01-06  Martin Sebor  <msebor@redhat.com>
30560         PR tree-optimization/83640
30561         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30562         subtracting negative offset from size.
30563         (builtin_access::overlap): Adjust offset bounds of the access to fall
30564         within the size of the object if possible.
30566 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30568         PR rtl-optimization/83699
30569         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30570         extract_bit_field_as_subreg to cases in which the extracted
30571         value is also a vector.
30573         * lra-constraints.c (process_alt_operands): Test for the equivalence
30574         substitutions when detecting a possible reload cycle.
30576 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30578         PR debug/83480
30579         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30580         by default if flag_selective_schedling{,2}.  Formatting fixes.
30582         PR rtl-optimization/83682
30583         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30584         if it has non-VECTOR_MODE element mode.
30585         (vec_duplicate_p): Likewise.
30587         PR middle-end/83694
30588         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30589         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30591 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30593         PR target/83604
30594         * config/i386/i386-builtin.def
30595         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30596         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30597         Require also OPTION_MASK_ISA_AVX512F in addition to
30598         OPTION_MASK_ISA_GFNI.
30599         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30600         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30601         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30602         to OPTION_MASK_ISA_GFNI.
30603         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30604         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30605         OPTION_MASK_ISA_AVX512BW.
30606         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30607         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30608         addition to OPTION_MASK_ISA_GFNI.
30609         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30610         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30611         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30612         to OPTION_MASK_ISA_GFNI.
30613         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30614         a requirement for all ISAs rather than any of them with a few
30615         exceptions.
30616         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30617         processing.
30618         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30619         bitmasks to be enabled with 3 exceptions, instead of requiring any
30620         enabled ISA with lots of exceptions.
30621         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30622         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30623         Change avx512bw in isa attribute to avx512f.
30624         * config/i386/sgxintrin.h: Add license boilerplate.
30625         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30626         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30627         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30628         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30629         defined.
30630         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30631         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30632         temporarily sse2 rather than sse if not enabled already.
30634         PR target/83604
30635         * config/i386/sse.md (VI248_VLBW): Rename to ...
30636         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30637         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30638         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30639         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30640         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30641         mode iterator instead of VI248_VLBW.
30643 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30645         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30646         (record_modified): Skip clobbers; add debug output.
30647         (param_change_prob): Use sreal frequencies.
30649 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30651         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30652         punt for user-aligned variables.
30654 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30656         * tree-chrec.c (chrec_contains_symbols): Return true for
30657         POLY_INT_CST.
30659 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30661         PR target/82439
30662         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30663         of (x|y) == x for BICS pattern.
30665 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30667         PR tree-optimization/83605
30668         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30669         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30670         can throw.
30672 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30674         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30675         * config/epiphany/rtems.h: New file.
30677 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30678             Uros Bizjak  <ubizjak@gmail.com>
30680         PR target/83554
30681         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30682         QIreg_operand instead of register_operand predicate.
30683         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30684         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30685         comments instead of -fmitigate[-_]rop.
30687 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30689         PR bootstrap/81926
30690         * cgraphunit.c (symbol_table::compile): Switch to text_section
30691         before calling assembly_start debug hook.
30692         * run-rtl-passes.c (run_rtl_passes): Likewise.
30693         Include output.h.
30695 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30697         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30698         range_int_cst_p rather than !symbolic_range_p before calling
30699         extract_range_from_multiplicative_op_1.
30701 2018-01-04  Jeff Law  <law@redhat.com>
30703         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30704         redundant test in assertion.
30706 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30708         * doc/rtl.texi: Document machine_mode wrapper classes.
30710 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30712         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30713         using tree_to_uhwi.
30715 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30717         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30718         the VEC_PERM_EXPR fold to fail.
30720 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30722         PR debug/83585
30723         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30724         to switched_sections.
30726 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30728         PR target/83680
30729         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30730         test for d.testing.
30732 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30734         PR target/83387
30735         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30736         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30738 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30740         PR debug/83666
30741         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30742         is BLKmode and bitpos not zero or mode change is needed.
30744 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30746         PR target/83675
30747         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30748         TARGET_VIS2.
30750 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30752         PR target/83628
30753         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30754         instead of MULT rtx.  Update all corresponding splitters.
30755         (*saddl_se): Ditto.
30756         (*ssub<modesuffix>): Ditto.
30757         (*ssubl_se): Ditto.
30758         (*cmp_sadd_di): Update split patterns.
30759         (*cmp_sadd_si): Ditto.
30760         (*cmp_sadd_sidi): Ditto.
30761         (*cmp_ssub_di): Ditto.
30762         (*cmp_ssub_si): Ditto.
30763         (*cmp_ssub_sidi): Ditto.
30764         * config/alpha/predicates.md (const23_operand): New predicate.
30765         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30766         Look for ASHIFT, not MULT inner operand.
30767         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30769 2018-01-04  Martin Liska  <mliska@suse.cz>
30771         PR gcov-profile/83669
30772         * gcov.c (output_intermediate_file): Add version to intermediate
30773         gcov file.
30774         * doc/gcov.texi: Document new field 'version' in intermediate
30775         file format. Fix location of '-k' option of gcov command.
30777 2018-01-04  Martin Liska  <mliska@suse.cz>
30779         PR ipa/82352
30780         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30782 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30784         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30786 2018-01-03  Martin Sebor  <msebor@redhat.com>
30788         PR tree-optimization/83655
30789         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30790         checking calls with invalid arguments.
30792 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30794         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30795         (vectorizable_mask_load_store): Delete.
30796         (vectorizable_call): Return false for masked loads and stores.
30797         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30798         instead of gimple_assign_rhs1.
30799         (vectorizable_load): Handle IFN_MASK_LOAD.
30800         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30802 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30804         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30805         split out from..,
30806         (vectorizable_mask_load_store): ...here.
30807         (vectorizable_load): ...and here.
30809 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30811         * tree-vect-stmts.c (vect_build_all_ones_mask)
30812         (vect_build_zero_merge_argument): New functions, split out from...
30813         (vectorizable_load): ...here.
30815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30817         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30818         split out from...
30819         (vectorizable_mask_load_store): ...here.
30820         (vectorizable_store): ...and here.
30822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30824         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30825         split out from...
30826         (vectorizable_mask_load_store): ...here.
30828 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30830         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30831         (vect_model_store_cost): Take a vec_load_store_type instead of a
30832         vect_def_type.
30833         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30834         (vect_model_store_cost): Take a vec_load_store_type instead of a
30835         vect_def_type.
30836         (vectorizable_mask_load_store): Update accordingly.
30837         (vectorizable_store): Likewise.
30838         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30840 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30842         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30843         IFN_MASK_LOAD calls here rather than...
30844         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30846 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30847             Alan Hayward  <alan.hayward@arm.com>
30848             David Sherwood  <david.sherwood@arm.com>
30850         * expmed.c (extract_bit_field_1): For vector extracts,
30851         fall back to extract_bit_field_as_subreg if vec_extract
30852         isn't available.
30854 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30855             Alan Hayward  <alan.hayward@arm.com>
30856             David Sherwood  <david.sherwood@arm.com>
30858         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30859         they are variable or constant sized.
30860         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30861         slots for constant-sized data.
30863 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30864             Alan Hayward  <alan.hayward@arm.com>
30865             David Sherwood  <david.sherwood@arm.com>
30867         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30868         handling COND_EXPRs with boolean comparisons, try to find a better
30869         basis for the mask type than the boolean itself.
30871 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30873         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30874         is calculated and how it can be overridden.
30875         * genmodes.c (max_bitsize_mode_any_mode): New variable.
30876         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30877         if defined.
30878         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30879         if nonzero.
30881 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30882             Alan Hayward  <alan.hayward@arm.com>
30883             David Sherwood  <david.sherwood@arm.com>
30885         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30886         Remove the mode argument.
30887         (aarch64_simd_valid_immediate): Remove the mode and inverse
30888         arguments.
30889         * config/aarch64/iterators.md (bitsize): New iterator.
30890         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30891         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30892         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30893         aarch64_simd_valid_immediate.
30894         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30895         (aarch64_reg_or_bic_imm): Likewise.
30896         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30897         with an insn_type enum and msl with a modifier_type enum.
30898         Replace element_width with a scalar_mode.  Change the shift
30899         to unsigned int.  Add constructors for scalar_float_mode and
30900         scalar_int_mode elements.
30901         (aarch64_vect_float_const_representable_p): Delete.
30902         (aarch64_can_const_movi_rtx_p)
30903         (aarch64_simd_scalar_immediate_valid_for_move)
30904         (aarch64_simd_make_constant): Update call to
30905         aarch64_simd_valid_immediate.
30906         (aarch64_advsimd_valid_immediate_hs): New function.
30907         (aarch64_advsimd_valid_immediate): Likewise.
30908         (aarch64_simd_valid_immediate): Remove mode and inverse
30909         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
30910         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30911         and aarch64_float_const_representable_p on the result.
30912         (aarch64_output_simd_mov_immediate): Remove mode argument.
30913         Update call to aarch64_simd_valid_immediate and use of
30914         simd_immediate_info.
30915         (aarch64_output_scalar_simd_mov_immediate): Update call
30916         accordingly.
30918 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30919             Alan Hayward  <alan.hayward@arm.com>
30920             David Sherwood  <david.sherwood@arm.com>
30922         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30923         (mode_nunits): Likewise CONST_MODE_NUNITS.
30924         * machmode.def (ADJUST_NUNITS): Document.
30925         * genmodes.c (mode_data::need_nunits_adj): New field.
30926         (blank_mode): Update accordingly.
30927         (adj_nunits): New variable.
30928         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30929         parameter.
30930         (emit_mode_size_inline): Set need_bytesize_adj for all modes
30931         listed in adj_nunits.
30932         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30933         listed in adj_nunits.  Don't emit case statements for such modes.
30934         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30935         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
30936         nothing if adj_nunits is nonnull.
30937         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30938         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30939         (emit_mode_fbit): Update use of print_maybe_const_decl.
30940         (emit_move_size): Likewise.  Treat the array as non-const
30941         if adj_nunits.
30942         (emit_mode_adjustments): Handle adj_nunits.
30944 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30946         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30947         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30948         (VECTOR_MODES): Use it.
30949         (make_vector_modes): Take the prefix as an argument.
30951 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30952             Alan Hayward  <alan.hayward@arm.com>
30953             David Sherwood  <david.sherwood@arm.com>
30955         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30956         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30957         for MODE_VECTOR_BOOL.
30958         * machmode.def (VECTOR_BOOL_MODE): Document.
30959         * genmodes.c (VECTOR_BOOL_MODE): New macro.
30960         (make_vector_bool_mode): New function.
30961         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30962         MODE_VECTOR_BOOL.
30963         * lto-streamer-in.c (lto_input_mode_table): Likewise.
30964         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30965         Likewise.
30966         * stor-layout.c (int_mode_for_mode): Likewise.
30967         * tree.c (build_vector_type_for_mode): Likewise.
30968         * varasm.c (output_constant_pool_2): Likewise.
30969         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30970         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
30971         for MODE_VECTOR_BOOL.
30972         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30973         of mode class checks.
30974         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30975         instead of a list of mode class checks.
30976         (expand_vector_scalar_condition): Likewise.
30977         (type_for_widest_vector_mode): Handle BImode as an inner mode.
30979 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30980             Alan Hayward  <alan.hayward@arm.com>
30981             David Sherwood  <david.sherwood@arm.com>
30983         * machmode.h (mode_size): Change from unsigned short to
30984         poly_uint16_pod.
30985         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30986         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30987         or if measurement_type is not polynomial.
30988         (fixed_size_mode::includes_p): Check for constant-sized modes.
30989         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30990         return a poly_uint16 rather than an unsigned short.
30991         (emit_mode_size): Change the type of mode_size from unsigned short
30992         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
30993         (emit_mode_adjustments): Cope with polynomial vector sizes.
30994         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30995         for GET_MODE_SIZE.
30996         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30997         for GET_MODE_SIZE.
30998         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30999         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31000         * caller-save.c (setup_save_areas): Likewise.
31001         (replace_reg_with_saved_mem): Likewise.
31002         * calls.c (emit_library_call_value_1): Likewise.
31003         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31004         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31005         (gen_lowpart_for_combine): Likewise.
31006         * convert.c (convert_to_integer_1): Likewise.
31007         * cse.c (equiv_constant, cse_insn): Likewise.
31008         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31009         (cselib_subst_to_values): Likewise.
31010         * dce.c (word_dce_process_block): Likewise.
31011         * df-problems.c (df_word_lr_mark_ref): Likewise.
31012         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31013         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31014         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31015         (rtl_for_decl_location): Likewise.
31016         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31017         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31018         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31019         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31020         (expand_expr_real_1): Likewise.
31021         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31022         (pad_below): Likewise.
31023         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31024         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31025         * ira.c (get_subreg_tracking_sizes): Likewise.
31026         * ira-build.c (ira_create_allocno_objects): Likewise.
31027         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31028         (ira_sort_regnos_for_alter_reg): Likewise.
31029         * ira-costs.c (record_operand_costs): Likewise.
31030         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31031         (resolve_simple_move): Likewise.
31032         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31033         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31034         (lra_constraints): Likewise.
31035         (CONST_POOL_OK_P): Reject variable-sized modes.
31036         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31037         (add_pseudo_to_slot, lra_spill): Likewise.
31038         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31039         * optabs-query.c (get_best_extraction_insn): Likewise.
31040         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31041         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31042         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31043         * recog.c (offsettable_address_addr_space_p): Likewise.
31044         * regcprop.c (maybe_mode_change): Likewise.
31045         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31046         * regrename.c (build_def_use): Likewise.
31047         * regstat.c (dump_reg_info): Likewise.
31048         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31049         (find_reloads, find_reloads_subreg_address): Likewise.
31050         * reload1.c (eliminate_regs_1): Likewise.
31051         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31052         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31053         (simplify_binary_operation_1, simplify_subreg): Likewise.
31054         * targhooks.c (default_function_arg_padding): Likewise.
31055         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31056         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31057         (verify_gimple_assign_ternary): Likewise.
31058         * tree-inline.c (estimate_move_cost): Likewise.
31059         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31060         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31061         (get_address_cost_ainc): Likewise.
31062         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31063         (vect_supportable_dr_alignment): Likewise.
31064         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31065         (vectorizable_reduction): Likewise.
31066         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31067         (vectorizable_operation, vectorizable_load): Likewise.
31068         * tree.c (build_same_sized_truth_vector_type): Likewise.
31069         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31070         * var-tracking.c (emit_note_insn_var_location): Likewise.
31071         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31072         (ADDR_VEC_ALIGN): Likewise.
31074 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31075             Alan Hayward  <alan.hayward@arm.com>
31076             David Sherwood  <david.sherwood@arm.com>
31078         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31079         unsigned short.
31080         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31081         or if measurement_type is polynomial.
31082         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31083         * combine.c (make_extraction): Likewise.
31084         * dse.c (find_shift_sequence): Likewise.
31085         * dwarf2out.c (mem_loc_descriptor): Likewise.
31086         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31087         (extract_bit_field, extract_low_bits): Likewise.
31088         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31089         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31090         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31091         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31092         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31093         * reload.c (find_reloads): Likewise.
31094         * reload1.c (alter_reg): Likewise.
31095         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31096         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31097         * tree-if-conv.c (predicate_mem_writes): Likewise.
31098         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31099         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31100         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31101         * valtrack.c (dead_debug_insert_temp): Likewise.
31102         * varasm.c (mergeable_constant_section): Likewise.
31103         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31106             Alan Hayward  <alan.hayward@arm.com>
31107             David Sherwood  <david.sherwood@arm.com>
31109         * expr.c (expand_assignment): Cope with polynomial mode sizes
31110         when assigning to a CONCAT.
31112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31113             Alan Hayward  <alan.hayward@arm.com>
31114             David Sherwood  <david.sherwood@arm.com>
31116         * machmode.h (mode_precision): Change from unsigned short to
31117         poly_uint16_pod.
31118         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31119         short.
31120         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31121         or if measurement_type is not polynomial.
31122         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31123         in which the mode is already known to be a scalar_int_mode.
31124         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31125         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31126         initializer.
31127         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31128         for GET_MODE_PRECISION.
31129         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31130         for GET_MODE_PRECISION.
31131         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31132         as polynomial.
31133         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31134         (expand_field_assignment, make_extraction): Likewise.
31135         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31136         (get_last_value): Likewise.
31137         * convert.c (convert_to_integer_1): Likewise.
31138         * cse.c (cse_insn): Likewise.
31139         * expr.c (expand_expr_real_1): Likewise.
31140         * lra-constraints.c (simplify_operand_subreg): Likewise.
31141         * optabs-query.c (can_atomic_load_p): Likewise.
31142         * optabs.c (expand_atomic_load): Likewise.
31143         (expand_atomic_store): Likewise.
31144         * ree.c (combine_reaching_defs): Likewise.
31145         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31146         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31147         * tree.h (type_has_mode_precision_p): Likewise.
31148         * ubsan.c (instrument_si_overflow): Likewise.
31150 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31151             Alan Hayward  <alan.hayward@arm.com>
31152             David Sherwood  <david.sherwood@arm.com>
31154         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31155         polynomial numbers of units.
31156         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31157         (valid_vector_subparts_p): New function.
31158         (build_vector_type): Remove temporary shim and take the number
31159         of units as a poly_uint64 rather than an int.
31160         (build_opaque_vector_type): Take the number of units as a
31161         poly_uint64 rather than an int.
31162         * tree.c (build_vector_from_ctor): Handle polynomial
31163         TYPE_VECTOR_SUBPARTS.
31164         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31165         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31166         (build_vector_from_val): If the number of units is variable,
31167         use build_vec_duplicate_cst for constant operands and
31168         VEC_DUPLICATE_EXPR otherwise.
31169         (make_vector_type): Remove temporary is_constant ().
31170         (build_vector_type, build_opaque_vector_type): Take the number of
31171         units as a poly_uint64 rather than an int.
31172         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31173         VECTOR_CST_NELTS.
31174         * cfgexpand.c (expand_debug_expr): Likewise.
31175         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31176         (store_constructor, expand_expr_real_1): Likewise.
31177         (const_scalar_mask_from_tree): Likewise.
31178         * fold-const-call.c (fold_const_reduction): Likewise.
31179         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31180         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31181         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31182         (fold_relational_const): Likewise.
31183         (native_interpret_vector): Likewise.  Change the size from an
31184         int to an unsigned int.
31185         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31186         TYPE_VECTOR_SUBPARTS.
31187         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31188         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31189         duplicating a non-constant operand into a variable-length vector.
31190         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31191         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31192         * ipa-icf.c (sem_variable::equals): Likewise.
31193         * match.pd: Likewise.
31194         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31195         * print-tree.c (print_node): Likewise.
31196         * stor-layout.c (layout_type): Likewise.
31197         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31198         * tree-cfg.c (verify_gimple_comparison): Likewise.
31199         (verify_gimple_assign_binary): Likewise.
31200         (verify_gimple_assign_ternary): Likewise.
31201         (verify_gimple_assign_single): Likewise.
31202         * tree-pretty-print.c (dump_generic_node): Likewise.
31203         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31204         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31205         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31206         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31207         (vect_shift_permute_load_chain): Likewise.
31208         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31209         (expand_vector_condition, optimize_vector_constructor): Likewise.
31210         (lower_vec_perm, get_compute_type): Likewise.
31211         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31212         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31213         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31214         (vect_recog_mask_conversion_pattern): Likewise.
31215         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31216         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31217         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31218         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31219         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31220         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31221         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31222         (supportable_widening_operation): Likewise.
31223         (supportable_narrowing_operation): Likewise.
31224         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31225         Likewise.
31226         * varasm.c (output_constant): Likewise.
31228 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31229             Alan Hayward  <alan.hayward@arm.com>
31230             David Sherwood  <david.sherwood@arm.com>
31232         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31233         so that both the length == 3 and length != 3 cases set up their
31234         own permute vectors.  Add comments explaining why we know the
31235         number of elements is constant.
31236         (vect_permute_load_chain): Likewise.
31238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31239             Alan Hayward  <alan.hayward@arm.com>
31240             David Sherwood  <david.sherwood@arm.com>
31242         * machmode.h (mode_nunits): Change from unsigned char to
31243         poly_uint16_pod.
31244         (ONLY_FIXED_SIZE_MODES): New macro.
31245         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31246         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31247         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31248         New typedefs.
31249         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31250         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31251         or if measurement_type is not polynomial.
31252         * genmodes.c (ZERO_COEFFS): New macro.
31253         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31254         poly_uint16.
31255         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31256         Use ZERO_COEFFS when emitting initializers.
31257         * data-streamer.h (bp_pack_poly_value): New function.
31258         (bp_unpack_poly_value): Likewise.
31259         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31260         for GET_MODE_NUNITS.
31261         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31262         for GET_MODE_NUNITS.
31263         * tree.c (make_vector_type): Remove temporary shim and make
31264         the real function take the number of units as a poly_uint64
31265         rather than an int.
31266         (build_vector_type_for_mode): Handle polynomial nunits.
31267         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31268         * emit-rtl.c (const_vec_series_p_1): Likewise.
31269         (gen_rtx_CONST_VECTOR): Likewise.
31270         * fold-const.c (test_vec_duplicate_folding): Likewise.
31271         * genrecog.c (validate_pattern): Likewise.
31272         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31273         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31274         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31275         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31276         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31277         * rtlanal.c (subreg_get_info): Likewise.
31278         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31279         (vect_grouped_load_supported): Likewise.
31280         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31281         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31282         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31283         (simplify_const_unary_operation, simplify_binary_operation_1)
31284         (simplify_const_binary_operation, simplify_ternary_operation)
31285         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31286         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31287         instead of CONST_VECTOR_NUNITS.
31288         * varasm.c (output_constant_pool_2): Likewise.
31289         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31290         explicit-encoded elements in the XVEC for variable-length vectors.
31292 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31294         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31297             Alan Hayward  <alan.hayward@arm.com>
31298             David Sherwood  <david.sherwood@arm.com>
31300         * coretypes.h (fixed_size_mode): Declare.
31301         (fixed_size_mode_pod): New typedef.
31302         * builtins.h (target_builtins::x_apply_args_mode)
31303         (target_builtins::x_apply_result_mode): Change type to
31304         fixed_size_mode_pod.
31305         * builtins.c (apply_args_size, apply_result_size, result_vector)
31306         (expand_builtin_apply_args_1, expand_builtin_apply)
31307         (expand_builtin_return): Update accordingly.
31309 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31311         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31312         * cselib.c (cselib_hash_rtx): Likewise.
31313         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31314         CONST_VECTOR encoding.
31316 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31317             Jeff Law  <law@redhat.com>
31319         PR target/83641
31320         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31321         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31322         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31323         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31325         PR target/83641
31326         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31327         explicitly probe *sp in a noreturn function if there were any callee
31328         register saves or frame pointer is needed.
31330 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31332         PR debug/83621
31333         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31334         BLKmode for ternary, binary or unary expressions.
31336         PR debug/83645
31337         * var-tracking.c (delete_vta_debug_insn): New inline function.
31338         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31339         insns from get_insns () to NULL instead of each bb separately.
31340         Use delete_vta_debug_insn.  No longer static.
31341         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31342         delete_vta_debug_insns callers.
31343         * rtl.h (delete_vta_debug_insns): Declare.
31344         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31345         instead of variable_tracking_main.
31347 2018-01-03  Martin Sebor  <msebor@redhat.com>
31349         PR tree-optimization/83603
31350         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31351         arguments past the endof the argument list in functions declared
31352         without a prototype.
31353         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31354         Avoid checking when arguments are null.
31356 2018-01-03  Martin Sebor  <msebor@redhat.com>
31358         PR c/83559
31359         * doc/extend.texi (attribute const): Fix a typo.
31360         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31361         issuing -Wsuggest-attribute for void functions.
31363 2018-01-03  Martin Sebor  <msebor@redhat.com>
31365         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31366         offset_int::from instead of wide_int::to_shwi.
31367         (maybe_diag_overlap): Remove assertion.
31368         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31369         * gimple-ssa-sprintf.c (format_directive): Same.
31370         (parse_directive): Same.
31371         (sprintf_dom_walker::compute_format_length): Same.
31372         (try_substitute_return_value): Same.
31374 2018-01-03  Jeff Law  <law@redhat.com>
31376         PR middle-end/83654
31377         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31378         non-constant residual for zero at runtime and avoid probing in
31379         that case.  Reorganize code for trailing problem to mirror handling
31380         of the residual.
31382 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31384         PR tree-optimization/83501
31385         * tree-ssa-strlen.c (get_string_cst): New.
31386         (handle_char_store): Call get_string_cst.
31388 2018-01-03  Martin Liska  <mliska@suse.cz>
31390         PR tree-optimization/83593
31391         * tree-ssa-strlen.c: Include tree-cfg.h.
31392         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31393         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31394         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31395         to false.
31396         (strlen_dom_walker::before_dom_children): Call
31397         gimple_purge_dead_eh_edges. Dump tranformation with details
31398         dump flags.
31399         (strlen_dom_walker::before_dom_children): Update call by adding
31400         new argument cleanup_eh.
31401         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31403 2018-01-03  Martin Liska  <mliska@suse.cz>
31405         PR ipa/83549
31406         * cif-code.def (VARIADIC_THUNK): New enum value.
31407         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31408         thunks.
31410 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31412         * sse.md (mov<mode>_internal): Tighten condition for when to use
31413         vmovdqu<ssescalarsize> for TI and OI modes.
31415 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31417         Update copyright years.
31419 2018-01-03  Martin Liska  <mliska@suse.cz>
31421         PR ipa/83594
31422         * ipa-visibility.c (function_and_variable_visibility): Skip
31423         functions with noipa attribure.
31425 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31427         * gcc.c (process_command): Update copyright notice dates.
31428         * gcov-dump.c (print_version): Ditto.
31429         * gcov.c (print_version): Ditto.
31430         * gcov-tool.c (print_version): Ditto.
31431         * gengtype.c (create_file): Ditto.
31432         * doc/cpp.texi: Bump @copying's copyright year.
31433         * doc/cppinternals.texi: Ditto.
31434         * doc/gcc.texi: Ditto.
31435         * doc/gccint.texi: Ditto.
31436         * doc/gcov.texi: Ditto.
31437         * doc/install.texi: Ditto.
31438         * doc/invoke.texi: Ditto.
31440 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31442         * vector-builder.h (vector_builder::m_full_nelts): Change from
31443         unsigned int to poly_uint64.
31444         (vector_builder::full_nelts): Update prototype accordingly.
31445         (vector_builder::new_vector): Likewise.
31446         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31447         (vector_builder::operator ==): Likewise.
31448         (vector_builder::finalize): Likewise.
31449         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31450         Take the number of elements as a poly_uint64 rather than an
31451         unsigned int.
31452         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31453         from unsigned int to poly_uint64.
31454         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31455         (vec_perm_indices::new_vector): Likewise.
31456         (vec_perm_indices::length): Likewise.
31457         (vec_perm_indices::nelts_per_input): Likewise.
31458         (vec_perm_indices::input_nelts): Likewise.
31459         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31460         number of elements per input as a poly_uint64 rather than an
31461         unsigned int.  Use the original encoding for variable-length
31462         vectors, rather than clamping each individual element.
31463         For the second and subsequent elements in each pattern,
31464         clamp the step and base before clamping their sum.
31465         (vec_perm_indices::series_p): Handle polynomial element counts.
31466         (vec_perm_indices::all_in_range_p): Likewise.
31467         (vec_perm_indices_to_tree): Likewise.
31468         (vec_perm_indices_to_rtx): Likewise.
31469         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31470         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31471         (tree_vector_builder::new_binary_operation): Handle polynomial
31472         element counts.  Return false if we need to know the number
31473         of elements at compile time.
31474         * fold-const.c (fold_vec_perm): Punt if the number of elements
31475         isn't known at compile time.
31477 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31479         * vec-perm-indices.h (vec_perm_builder): Change element type
31480         from HOST_WIDE_INT to poly_int64.
31481         (vec_perm_indices::element_type): Update accordingly.
31482         (vec_perm_indices::clamp): Handle polynomial element_types.
31483         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31484         (vec_perm_indices::all_in_range_p): Likewise.
31485         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31486         than shwi trees.
31487         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31488         polynomial vec_perm_indices element types.
31489         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31490         * fold-const.c (fold_vec_perm): Likewise.
31491         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31492         * tree-vect-generic.c (lower_vec_perm): Likewise.
31493         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31494         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31495         element type to HOST_WIDE_INT.
31497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31498             Alan Hayward  <alan.hayward@arm.com>
31499             David Sherwood  <david.sherwood@arm.com>
31501         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31502         rather than an int.  Use plus_constant.
31503         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31504         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31506 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31507             Alan Hayward  <alan.hayward@arm.com>
31508             David Sherwood  <david.sherwood@arm.com>
31510         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31511         a HOST_WIDE_INT to a poly_int64.
31513 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31514             Alan Hayward  <alan.hayward@arm.com>
31515             David Sherwood  <david.sherwood@arm.com>
31517         * calls.c (load_register_parameters): Cope with polynomial
31518         mode sizes.  Require a constant size for BLKmode parameters
31519         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31520         forces a parameter to be padded at the lsb end in order to
31521         fill a complete number of words, require the parameter size
31522         to be ordered wrt UNITS_PER_WORD.
31524 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31525             Alan Hayward  <alan.hayward@arm.com>
31526             David Sherwood  <david.sherwood@arm.com>
31528         * reload1.c (spill_stack_slot_width): Change element type
31529         from unsigned int to poly_uint64_pod.
31530         (alter_reg): Treat mode sizes as polynomial.
31532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31533             Alan Hayward  <alan.hayward@arm.com>
31534             David Sherwood  <david.sherwood@arm.com>
31536         * reload.c (complex_word_subreg_p): New function.
31537         (reload_inner_reg_of_subreg, push_reload): Use it.
31539 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31540             Alan Hayward  <alan.hayward@arm.com>
31541             David Sherwood  <david.sherwood@arm.com>
31543         * lra-constraints.c (process_alt_operands): Reject matched
31544         operands whose sizes aren't ordered.
31545         (match_reload): Refer to this check here.
31547 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31548             Alan Hayward  <alan.hayward@arm.com>
31549             David Sherwood  <david.sherwood@arm.com>
31551         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31552         that the mode size is in the set {1, 2, 4, 8, 16}.
31554 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31555             Alan Hayward  <alan.hayward@arm.com>
31556             David Sherwood  <david.sherwood@arm.com>
31558         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31559         Use plus_constant instead of gen_rtx_PLUS.
31561 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31562             Alan Hayward  <alan.hayward@arm.com>
31563             David Sherwood  <david.sherwood@arm.com>
31565         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31566         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31567         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31568         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31569         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31570         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31571         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31572         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31573         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31574         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31575         a poly_int64.
31576         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31577         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31578         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31579         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31580         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31581         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31582         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31583         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31584         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31585         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31586         function.
31587         * expr.c (emit_move_resolve_push): Treat the input and result
31588         of PUSH_ROUNDING as a poly_int64.
31589         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31590         (emit_push_insn): Likewise.
31591         * lra-eliminations.c (mark_not_eliminable): Likewise.
31592         * recog.c (push_operand): Likewise.
31593         * reload1.c (elimination_effects): Likewise.
31594         * rtlanal.c (nonzero_bits1): Likewise.
31595         * calls.c (store_one_arg): Likewise.  Require the padding to be
31596         known at compile time.
31598 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31599             Alan Hayward  <alan.hayward@arm.com>
31600             David Sherwood  <david.sherwood@arm.com>
31602         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31603         Use plus_constant instead of gen_rtx_PLUS.
31605 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31606             Alan Hayward  <alan.hayward@arm.com>
31607             David Sherwood  <david.sherwood@arm.com>
31609         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31610         rather than an int.
31612 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31613             Alan Hayward  <alan.hayward@arm.com>
31614             David Sherwood  <david.sherwood@arm.com>
31616         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31617         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31618         via stack temporaries.  Treat the mode size as polynomial too.
31620 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31621             Alan Hayward  <alan.hayward@arm.com>
31622             David Sherwood  <david.sherwood@arm.com>
31624         * expr.c (expand_expr_real_2): When handling conversions involving
31625         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31626         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31627         as a poly_uint64 too.
31629 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31630             Alan Hayward  <alan.hayward@arm.com>
31631             David Sherwood  <david.sherwood@arm.com>
31633         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31635 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31636             Alan Hayward  <alan.hayward@arm.com>
31637             David Sherwood  <david.sherwood@arm.com>
31639         * combine.c (can_change_dest_mode): Handle polynomial
31640         REGMODE_NATURAL_SIZE.
31641         * expmed.c (store_bit_field_1): Likewise.
31642         * expr.c (store_constructor): Likewise.
31643         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31644         and polynomial REGMODE_NATURAL_SIZE.
31645         (gen_lowpart_common): Likewise.
31646         * reginfo.c (record_subregs_of_mode): Likewise.
31647         * rtlanal.c (read_modify_subreg_p): Likewise.
31649 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31650             Alan Hayward  <alan.hayward@arm.com>
31651             David Sherwood  <david.sherwood@arm.com>
31653         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31654         numbers of elements.
31656 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31657             Alan Hayward  <alan.hayward@arm.com>
31658             David Sherwood  <david.sherwood@arm.com>
31660         * match.pd: Cope with polynomial numbers of vector elements.
31662 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31663             Alan Hayward  <alan.hayward@arm.com>
31664             David Sherwood  <david.sherwood@arm.com>
31666         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31667         in a POINTER_PLUS_EXPR.
31669 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31670             Alan Hayward  <alan.hayward@arm.com>
31671             David Sherwood  <david.sherwood@arm.com>
31673         * omp-simd-clone.c (simd_clone_subparts): New function.
31674         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31675         (ipa_simd_modify_function_body): Likewise.
31677 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31678             Alan Hayward  <alan.hayward@arm.com>
31679             David Sherwood  <david.sherwood@arm.com>
31681         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31682         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31683         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31684         (expand_vector_condition, vector_element): Likewise.
31685         (subparts_gt): New function.
31686         (get_compute_type): Use subparts_gt.
31687         (count_type_subparts): Delete.
31688         (expand_vector_operations_1): Use subparts_gt instead of
31689         count_type_subparts.
31691 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31692             Alan Hayward  <alan.hayward@arm.com>
31693             David Sherwood  <david.sherwood@arm.com>
31695         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31696         (vect_compile_time_alias): ...this new function.  Do the calculation
31697         on poly_ints rather than trees.
31698         (vect_prune_runtime_alias_test_list): Update call accordingly.
31700 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31701             Alan Hayward  <alan.hayward@arm.com>
31702             David Sherwood  <david.sherwood@arm.com>
31704         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31705         numbers of units.
31706         (vect_schedule_slp_instance): Likewise.
31708 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31709             Alan Hayward  <alan.hayward@arm.com>
31710             David Sherwood  <david.sherwood@arm.com>
31712         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31713         constant and extern definitions for variable-length vectors.
31714         (vect_get_constant_vectors): Note that the number of units
31715         is known to be constant.
31717 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31718             Alan Hayward  <alan.hayward@arm.com>
31719             David Sherwood  <david.sherwood@arm.com>
31721         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31722         of units as polynomial.  Choose between WIDE and NARROW based
31723         on multiple_p.
31725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31726             Alan Hayward  <alan.hayward@arm.com>
31727             David Sherwood  <david.sherwood@arm.com>
31729         * tree-vect-stmts.c (simd_clone_subparts): New function.
31730         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31732 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31733             Alan Hayward  <alan.hayward@arm.com>
31734             David Sherwood  <david.sherwood@arm.com>
31736         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31737         vectors as polynomial.  Use build_index_vector for
31738         IFN_GOMP_SIMD_LANE.
31740 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31741             Alan Hayward  <alan.hayward@arm.com>
31742             David Sherwood  <david.sherwood@arm.com>
31744         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31745         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31746         for variable-length vectors.
31747         (vectorizable_mask_load_store): Treat the number of units as
31748         polynomial, asserting that it is constant if the condition has
31749         already been enforced.
31750         (vectorizable_store, vectorizable_load): Likewise.
31752 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31753             Alan Hayward  <alan.hayward@arm.com>
31754             David Sherwood  <david.sherwood@arm.com>
31756         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31757         of units as polynomial.  Punt if we can't tell at compile time
31758         which vector contains the final result.
31760 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31761             Alan Hayward  <alan.hayward@arm.com>
31762             David Sherwood  <david.sherwood@arm.com>
31764         * tree-vect-loop.c (vectorizable_induction): Treat the number
31765         of units as polynomial.  Punt on SLP inductions.  Use an integer
31766         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31767         cast of such a series for variable-length floating-point
31768         reductions.
31770 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31771             Alan Hayward  <alan.hayward@arm.com>
31772             David Sherwood  <david.sherwood@arm.com>
31774         * tree.h (build_index_vector): Declare.
31775         * tree.c (build_index_vector): New function.
31776         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31777         of units as polynomial, forcibly converting it to a constant if
31778         vectorizable_reduction has already enforced the condition.
31779         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31780         to create a {1,2,3,...} vector.
31781         (vectorizable_reduction): Treat the number of units as polynomial.
31782         Choose vectype_in based on the largest scalar element size rather
31783         than the smallest number of units.  Enforce the restrictions
31784         relied on above.
31786 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31787             Alan Hayward  <alan.hayward@arm.com>
31788             David Sherwood  <david.sherwood@arm.com>
31790         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31791         number of units as polynomial.
31793 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31794             Alan Hayward  <alan.hayward@arm.com>
31795             David Sherwood  <david.sherwood@arm.com>
31797         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31798         * target.def (autovectorize_vector_sizes): Return the vector sizes
31799         by pointer, using vector_sizes rather than a bitmask.
31800         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31801         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31802         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31803         Likewise.
31804         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31805         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31806         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31807         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31808         * omp-general.c (omp_max_vf): Likewise.
31809         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31810         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31811         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31812         * tree-vect-slp.c (vect_slp_bb): Likewise.
31813         * doc/tm.texi: Regenerate.
31814         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31815         to a poly_uint64.
31816         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31817         the vector size as a poly_uint64 rather than an unsigned int.
31818         (current_vector_size): Change from an unsigned int to a poly_uint64.
31819         (get_vectype_for_scalar_type): Update accordingly.
31820         * tree.h (build_truth_vector_type): Take the size and number of
31821         units as a poly_uint64 rather than an unsigned int.
31822         (build_vector_type): Add a temporary overload that takes
31823         the number of units as a poly_uint64 rather than an unsigned int.
31824         * tree.c (make_vector_type): Likewise.
31825         (build_truth_vector_type): Take the number of units as a poly_uint64
31826         rather than an unsigned int.
31828 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31829             Alan Hayward  <alan.hayward@arm.com>
31830             David Sherwood  <david.sherwood@arm.com>
31832         * target.def (get_mask_mode): Take the number of units and length
31833         as poly_uint64s rather than unsigned ints.
31834         * targhooks.h (default_get_mask_mode): Update accordingly.
31835         * targhooks.c (default_get_mask_mode): Likewise.
31836         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31837         * doc/tm.texi: Regenerate.
31839 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31840             Alan Hayward  <alan.hayward@arm.com>
31841             David Sherwood  <david.sherwood@arm.com>
31843         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31844         * omp-general.c (omp_max_vf): Likewise.
31845         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31846         (expand_omp_simd): Handle polynomial safelen.
31847         * omp-low.c (omplow_simd_context): Add a default constructor.
31848         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31849         (lower_rec_simd_input_clauses): Update accordingly.
31850         (lower_rec_input_clauses): Likewise.
31852 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31853             Alan Hayward  <alan.hayward@arm.com>
31854             David Sherwood  <david.sherwood@arm.com>
31856         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31857         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31858         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31859         (vect_analyze_slp_cost): Likewise.
31860         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31861         (vect_model_load_cost): Likewise.
31863 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31864             Alan Hayward  <alan.hayward@arm.com>
31865             David Sherwood  <david.sherwood@arm.com>
31867         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31868         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31869         from an unsigned int * to a poly_uint64_pod *.
31870         (calculate_unrolling_factor): New function.
31871         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
31873 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31874             Alan Hayward  <alan.hayward@arm.com>
31875             David Sherwood  <david.sherwood@arm.com>
31877         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31878         from an unsigned int to a poly_uint64.
31879         (_loop_vec_info::slp_unrolling_factor): Likewise.
31880         (_loop_vec_info::vectorization_factor): Change from an int
31881         to a poly_uint64.
31882         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31883         (vect_get_num_vectors): New function.
31884         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31885         (vect_get_num_copies): Use vect_get_num_vectors.
31886         (vect_analyze_data_ref_dependences): Change max_vf from an int *
31887         to an unsigned int *.
31888         (vect_analyze_data_refs): Change min_vf from an int * to a
31889         poly_uint64 *.
31890         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31891         than an unsigned HOST_WIDE_INT.
31892         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31893         (vect_analyze_data_ref_dependence): Change max_vf from an int *
31894         to an unsigned int *.
31895         (vect_analyze_data_ref_dependences): Likewise.
31896         (vect_compute_data_ref_alignment): Handle polynomial vf.
31897         (vect_enhance_data_refs_alignment): Likewise.
31898         (vect_prune_runtime_alias_test_list): Likewise.
31899         (vect_shift_permute_load_chain): Likewise.
31900         (vect_supportable_dr_alignment): Likewise.
31901         (dependence_distance_ge_vf): Take the vectorization factor as a
31902         poly_uint64 rather than an unsigned HOST_WIDE_INT.
31903         (vect_analyze_data_refs): Change min_vf from an int * to a
31904         poly_uint64 *.
31905         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31906         vfm1 as a poly_uint64 rather than an int.  Make the same change
31907         for the returned bound_scalar.
31908         (vect_gen_vector_loop_niters): Handle polynomial vf.
31909         (vect_do_peeling): Likewise.  Update call to
31910         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31911         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31912         be constant.
31913         * tree-vect-loop.c (vect_determine_vectorization_factor)
31914         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31915         (vect_get_known_peeling_cost): Likewise.
31916         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31917         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31918         (vect_transform_loop): Likewise.  Use the lowest possible VF when
31919         updating the upper bounds of the loop.
31920         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
31921         rather than an int.
31922         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31923         polynomial unroll factors.
31924         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31925         (vect_make_slp_decision): Likewise.
31926         (vect_supported_load_permutation_p): Likewise, and polynomial
31927         vf too.
31928         (vect_analyze_slp_cost): Handle polynomial vf.
31929         (vect_slp_analyze_node_operations): Likewise.
31930         (vect_slp_analyze_bb_1): Likewise.
31931         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31932         than an unsigned HOST_WIDE_INT.
31933         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31934         (vectorizable_load): Handle polynomial vf.
31935         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31936         a poly_uint64.
31937         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31939 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31940             Alan Hayward  <alan.hayward@arm.com>
31941             David Sherwood  <david.sherwood@arm.com>
31943         * match.pd: Handle bit operations involving three constants
31944         and try to fold one pair.
31946 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31948         * tree-vect-loop-manip.c: Include gimple-fold.h.
31949         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31950         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
31951         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31952         Add a path that uses a step of VF instead of 1, but disable it
31953         for now.
31954         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31955         and niters_no_overflow parameters.  Update calls to
31956         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31957         Create a new SSA name if the latter choses to use a ste other
31958         than zero, and return it via niters_vector_mult_vf_var.
31959         * tree-vect-loop.c (vect_transform_loop): Update calls to
31960         vect_do_peeling, vect_gen_vector_loop_niters and
31961         slpeel_make_loop_iterate_ntimes.
31962         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31963         (vect_gen_vector_loop_niters): Update declarations after above changes.
31965 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
31967         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31968         128-bit round to integer instructions.
31969         (ceil<mode>2): Likewise.
31970         (btrunc<mode>2): Likewise.
31971         (round<mode>2): Likewise.
31973 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31975         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31976         unaligned VSX load/store on P8/P9.
31977         (expand_block_clear): Allow the use of unaligned VSX
31978         load/store on P8/P9.
31980 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31982         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31983         New function.
31984         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31985         swap associated with both a load and a store.
31987 2018-01-02  Andrew Waterman  <andrew@sifive.com>
31989         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31990         * config/riscv/riscv.md (clear_cache): Use it.
31992 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
31994         * web.c: Remove out-of-date comment.
31996 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31998         * expr.c (fixup_args_size_notes): Check that any existing
31999         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32000         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32001         (emit_single_push_insn): ...here.
32003 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32005         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32006         (const_vector_encoded_nelts): New function.
32007         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32008         (const_vector_int_elt, const_vector_elt): Declare.
32009         * emit-rtl.c (const_vector_int_elt_1): New function.
32010         (const_vector_elt): Likewise.
32011         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32012         of CONST_VECTOR_ELT.
32014 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32016         * expr.c: Include rtx-vector-builder.h.
32017         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32018         directly on the tree encoding.
32019         (const_vector_from_tree): Likewise.
32020         * optabs.c: Include rtx-vector-builder.h.
32021         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32022         sequence of "u" values.
32023         * vec-perm-indices.c: Include rtx-vector-builder.h.
32024         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32025         directly on the vec_perm_indices encoding.
32027 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32029         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32030         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32031         * rtx-vector-builder.h: New file.
32032         * rtx-vector-builder.c: Likewise.
32033         * rtl.h (rtx_def::u2): Add a const_vector field.
32034         (CONST_VECTOR_NPATTERNS): New macro.
32035         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32036         (CONST_VECTOR_DUPLICATE_P): Likewise.
32037         (CONST_VECTOR_STEPPED_P): Likewise.
32038         (CONST_VECTOR_ENCODED_ELT): Likewise.
32039         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32040         (unwrap_const_vec_duplicate): Likewise.
32041         (const_vec_series_p): Check for a non-duplicated vector encoding.
32042         Say that the function only returns true for integer vectors.
32043         * emit-rtl.c: Include rtx-vector-builder.h.
32044         (gen_const_vec_duplicate_1): Delete.
32045         (gen_const_vector): Call gen_const_vec_duplicate instead of
32046         gen_const_vec_duplicate_1.
32047         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32048         (gen_const_vec_duplicate): Use rtx_vector_builder.
32049         (gen_const_vec_series): Likewise.
32050         (gen_rtx_CONST_VECTOR): Likewise.
32051         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32052         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32053         Build a new vector rather than modifying a CONST_VECTOR in-place.
32054         (handle_special_swappables): Update call accordingly.
32055         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32056         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32057         Build a new vector rather than modifying a CONST_VECTOR in-place.
32058         (handle_special_swappables): Update call accordingly.
32060 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32062         * simplify-rtx.c (simplify_const_binary_operation): Use
32063         CONST_VECTOR_ELT instead of XVECEXP.
32065 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32067         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32068         the selector elements to be different from the data elements
32069         if the selector is a VECTOR_CST.
32070         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32071         ssizetype for the selector.
32073 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32075         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32076         before testing each element individually.
32077         * tree-vect-generic.c (lower_vec_perm): Likewise.
32079 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32081         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32082         * selftest-run-tests.c (selftest::run_tests): Call it.
32083         * vector-builder.h (vector_builder::operator ==): New function.
32084         (vector_builder::operator !=): Likewise.
32085         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32086         (vec_perm_indices::all_from_input_p): New function.
32087         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32088         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32089         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32090         instead of reading the VECTOR_CST directly.  Detect whether both
32091         vector inputs are the same before constructing the vec_perm_indices,
32092         and update the number of inputs argument accordingly.  Use the
32093         utility functions added above.  Only construct sel2 if we need to.
32095 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32097         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32098         the broadcast of the low byte.
32099         (expand_mult_highpart): Use an explicit encoding for the permutes.
32100         * optabs-query.c (can_mult_highpart_p): Likewise.
32101         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32102         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32103         (vectorizable_bswap): Likewise.
32104         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32105         explicit encoding for the power-of-2 permutes.
32106         (vect_permute_store_chain): Likewise.
32107         (vect_grouped_load_supported): Likewise.
32108         (vect_permute_load_chain): Likewise.
32110 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32112         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32113         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32114         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32115         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32116         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32117         (vect_gen_perm_mask_any): Likewise.
32119 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32121         * int-vector-builder.h: New file.
32122         * vec-perm-indices.h: Include int-vector-builder.h.
32123         (vec_perm_indices): Redefine as an int_vector_builder.
32124         (auto_vec_perm_indices): Delete.
32125         (vec_perm_builder): Redefine as a stand-alone class.
32126         (vec_perm_indices::vec_perm_indices): New function.
32127         (vec_perm_indices::clamp): Likewise.
32128         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32129         (vec_perm_indices::new_vector): New function.
32130         (vec_perm_indices::new_expanded_vector): Update for new
32131         vec_perm_indices class.
32132         (vec_perm_indices::rotate_inputs): New function.
32133         (vec_perm_indices::all_in_range_p): Operate directly on the
32134         encoded form, without computing elided elements.
32135         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32136         encoding.  Update for new vec_perm_indices class.
32137         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32138         the given vec_perm_builder.
32139         (expand_vec_perm_var): Update vec_perm_builder constructor.
32140         (expand_mult_highpart): Use vec_perm_builder instead of
32141         auto_vec_perm_indices.
32142         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32143         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32144         or double series encoding as appropriate.
32145         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32146         vec_perm_indices instead of auto_vec_perm_indices.
32147         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32148         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32149         (vect_permute_store_chain): Likewise.
32150         (vect_grouped_load_supported): Likewise.
32151         (vect_permute_load_chain): Likewise.
32152         (vect_shift_permute_load_chain): Likewise.
32153         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32154         (vect_transform_slp_perm_load): Likewise.
32155         (vect_schedule_slp_instance): Likewise.
32156         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32157         (vectorizable_mask_load_store): Likewise.
32158         (vectorizable_bswap): Likewise.
32159         (vectorizable_store): Likewise.
32160         (vectorizable_load): Likewise.
32161         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32162         vec_perm_indices instead of auto_vec_perm_indices.  Use
32163         tree_to_vec_perm_builder to read the vector from a tree.
32164         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32165         vec_perm_builder instead of a vec_perm_indices.
32166         (have_whole_vector_shift): Use vec_perm_builder and
32167         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32168         truncation to calc_vec_perm_mask_for_shift.
32169         (vect_create_epilog_for_reduction): Likewise.
32170         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32171         from auto_vec_perm_indices to vec_perm_indices.
32172         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32173         instead of changing individual elements.
32174         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32175         the vector in d.perm.
32176         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32177         from auto_vec_perm_indices to vec_perm_indices.
32178         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32179         instead of changing individual elements.
32180         (arm_vectorize_vec_perm_const): Use new_vector to install
32181         the vector in d.perm.
32182         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32183         Update vec_perm_builder constructor.
32184         (rs6000_expand_interleave): Likewise.
32185         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32186         (rs6000_expand_interleave): Likewise.
32188 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32190         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32191         to qimode could truncate the indices.
32192         * optabs.c (expand_vec_perm_var): Likewise.
32194 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32196         * Makefile.in (OBJS): Add vec-perm-indices.o.
32197         * vec-perm-indices.h: New file.
32198         * vec-perm-indices.c: Likewise.
32199         * target.h (vec_perm_indices): Replace with a forward class
32200         declaration.
32201         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32202         * optabs.h: Include vec-perm-indices.h.
32203         (expand_vec_perm): Delete.
32204         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32205         (expand_vec_perm_const): Declare.
32206         * target.def (vec_perm_const_ok): Replace with...
32207         (vec_perm_const): ...this new hook.
32208         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32209         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32210         * doc/tm.texi: Regenerate.
32211         * optabs.def (vec_perm_const): Delete.
32212         * doc/md.texi (vec_perm_const): Likewise.
32213         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32214         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32215         expand_vec_perm for constant permutation vectors.  Assert that
32216         the mode of variable permutation vectors is the integer equivalent
32217         of the mode that is being permuted.
32218         * optabs-query.h (selector_fits_mode_p): Declare.
32219         * optabs-query.c: Include vec-perm-indices.h.
32220         (selector_fits_mode_p): New function.
32221         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32222         is defined, instead of checking whether the vec_perm_const_optab
32223         exists.  Use targetm.vectorize.vec_perm_const instead of
32224         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32225         fit in the vector mode before using a variable permute.
32226         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32227         vec_perm_indices instead of an rtx.
32228         (expand_vec_perm): Replace with...
32229         (expand_vec_perm_const): ...this new function.  Take the selector
32230         as a vec_perm_indices rather than an rtx.  Also take the mode of
32231         the selector.  Update call to shift_amt_for_vec_perm_mask.
32232         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32233         Use vec_perm_indices::new_expanded_vector to expand the original
32234         selector into bytes.  Check whether the indices fit in the vector
32235         mode before using a variable permute.
32236         (expand_vec_perm_var): Make global.
32237         (expand_mult_highpart): Use expand_vec_perm_const.
32238         * fold-const.c: Includes vec-perm-indices.h.
32239         * tree-ssa-forwprop.c: Likewise.
32240         * tree-vect-data-refs.c: Likewise.
32241         * tree-vect-generic.c: Likewise.
32242         * tree-vect-loop.c: Likewise.
32243         * tree-vect-slp.c: Likewise.
32244         * tree-vect-stmts.c: Likewise.
32245         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32246         Delete.
32247         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32248         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32249         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32250         (aarch64_vectorize_vec_perm_const): ...this new function.
32251         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32252         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32253         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32254         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32255         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32256         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32257         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32258         into...
32259         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32260         check for NEON modes.
32261         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32262         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32263         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32264         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32265         into...
32266         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32267         the old VEC_PERM_CONST conditions.
32268         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32269         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32270         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32271         (ia64_vectorize_vec_perm_const_ok): Merge into...
32272         (ia64_vectorize_vec_perm_const): ...this new function.
32273         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32274         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32275         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32276         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32277         * config/mips/mips.c (mips_expand_vec_perm_const)
32278         (mips_vectorize_vec_perm_const_ok): Merge into...
32279         (mips_vectorize_vec_perm_const): ...this new function.
32280         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32281         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32282         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32283         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32284         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32285         (rs6000_expand_vec_perm_const): Delete.
32286         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32287         Delete.
32288         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32289         (altivec_expand_vec_perm_const_le): Take each operand individually.
32290         Operate on constant selectors rather than rtxes.
32291         (altivec_expand_vec_perm_const): Likewise.  Update call to
32292         altivec_expand_vec_perm_const_le.
32293         (rs6000_expand_vec_perm_const): Delete.
32294         (rs6000_vectorize_vec_perm_const_ok): Delete.
32295         (rs6000_vectorize_vec_perm_const): New function.
32296         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32297         an element count and rtx array.
32298         (rs6000_expand_extract_even): Update call accordingly.
32299         (rs6000_expand_interleave): Likewise.
32300         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32301         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32302         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32303         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32304         (rs6000_expand_vec_perm_const): Delete.
32305         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32306         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32307         (altivec_expand_vec_perm_const_le): Take each operand individually.
32308         Operate on constant selectors rather than rtxes.
32309         (altivec_expand_vec_perm_const): Likewise.  Update call to
32310         altivec_expand_vec_perm_const_le.
32311         (rs6000_expand_vec_perm_const): Delete.
32312         (rs6000_vectorize_vec_perm_const_ok): Delete.
32313         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32314         reference to the SPE evmerge intructions.
32315         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32316         an element count and rtx array.
32317         (rs6000_expand_extract_even): Update call accordingly.
32318         (rs6000_expand_interleave): Likewise.
32319         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32320         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32321         new function.
32322         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32324 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32326         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32327         vector mode and that that mode matches the mode of the data
32328         being permuted.
32329         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32330         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32331         directly using expand_vec_perm_1 when forcing selectors into
32332         registers.
32333         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32335 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32337         * optabs-query.h (can_vec_perm_p): Delete.
32338         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32339         * optabs-query.c (can_vec_perm_p): Split into...
32340         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32341         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32342         particular selector is valid.
32343         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32344         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32345         (vect_grouped_load_supported): Likewise.
32346         (vect_shift_permute_load_chain): Likewise.
32347         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32348         (vect_transform_slp_perm_load): Likewise.
32349         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32350         (vectorizable_bswap): Likewise.
32351         (vect_gen_perm_mask_checked): Likewise.
32352         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32353         implementations of variable permutation vectors into account
32354         when deciding which selector to use.
32355         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32356         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32357         with a false third argument.
32358         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32359         to test whether the constant selector is valid and can_vec_perm_var_p
32360         to test whether a variable selector is valid.
32362 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32364         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32365         * optabs-query.c (can_vec_perm_p): Likewise.
32366         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32367         instead of vec_perm_indices.
32368         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32369         (vect_gen_perm_mask_checked): Likewise,
32370         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32371         (vect_gen_perm_mask_checked): Likewise,
32373 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32375         * optabs-query.h (qimode_for_vec_perm): Declare.
32376         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32377         (qimode_for_vec_perm): ...this new function.
32378         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32380 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32382         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32383         does not have a conditional at the top.
32385 2018-01-02  Richard Biener  <rguenther@suse.de>
32387         * ipa-inline.c (big_speedup_p): Fix expression.
32389 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32391         PR target/81616
32392         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32393         for generic 4->6.
32395 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32397         PR target/81616
32398         Generic tuning.
32399         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32400         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32401         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32402         cond_taken_branch_cost 3->4.
32404 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32406         PR tree-optimization/83581
32407         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32408         TODO_cleanup_cfg if any changes have been made.
32410         PR middle-end/83608
32411         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32412         convert_modes if target mode has the right side, but different mode
32413         class.
32415         PR middle-end/83609
32416         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32417         last argument when extracting from CONCAT.  If either from_real or
32418         from_imag is NULL, use expansion through memory.  If result is not
32419         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32420         the parts directly to inner mode, if even that fails, use expansion
32421         through memory.
32423         PR middle-end/83623
32424         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32425         check for bswap in mode rather than HImode and use that in expand_unop
32426         too.
32428 Copyright (C) 2018 Free Software Foundation, Inc.
32430 Copying and distribution of this file, with or without modification,
32431 are permitted in any medium without royalty provided the copyright
32432 notice and this notice are preserved.