1 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5 for removed I387_MASK_PM entity.
7 2018-09-26 Jeff Law <law@redhat.com>
10 2018-09-26 Alexey Neyman <stilor@att.net>
12 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
13 headers are no longer pulled in by <isl/val.h>.
15 2018-09-26 Richard Biener <rguenther@suse.de>
18 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
19 or concrete instance DIE to the tree. Create abstract origin
20 attributes also for concrete instances.
22 2018-09-26 Alexey Neyman <stilor@att.net>
24 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
25 headers are no longer pulled in by <isl/val.h>.
27 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
29 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
30 Use new helper functions.
31 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
32 Use new helper functions.
33 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
34 aarch_mm_needs_release): New declarations.
35 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
36 aarch_mm_needs_release): New.
38 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
40 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
41 (arm32_output_mi_thunk): Deal with long calls.
43 2018-09-26 Richard Biener <rguenther@suse.de>
47 * tree-inline.c (expand_call_inline): When the location
48 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
49 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
50 the inserted BLOCK to make inlined_function_outer_scope_p
52 * dwarf2out.c (add_call_src_coords_attributes): Do not add
53 coords for reserved locations.
55 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
57 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
58 (*call_indirect_nonlocal_sysv<mode>): Ditto.
59 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
60 (*sibcall_nonlocal_sysv<mode>): Ditto.
61 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
63 (<bd>tf_<mode>): Ditto.
65 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
67 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
68 control string as a list of templates instead of as C code.
69 (*altivec_movti): Ditto.
70 * config/rs6000/darwin.md (movdf_low_di): Ditto.
72 2018-09-25 Jim Wilson <jimw@sifive.com>
74 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
75 when target symbol is weak.
77 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
80 * builtins.c (unterminated_array): Simplify.
81 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
82 where pointer arithmetic is safe.
84 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
87 * config/rs6000/altivec.md (altivec_vspltb): Use
88 const_0_to_15_operand instead of u5bit_cint_operand.
89 (*altivec_vspltb_internal): Ditto.
90 (altivec_vspltb_direct): Ditto.
91 (altivec_vsplth): Use const_0_to_7_operand instead of
93 (*altivec_vsplth_internal): Ditto.
94 (altivec_vsplth_direct): Ditto.
95 (altivec_vspltw): Use const_0_to_3_operand instead of
97 (*altivec_vspltw_internal): Ditto.
98 (altivec_vspltw_direct): Ditto.
99 (altivec_vspltsf): Ditto.
100 (*altivec_vspltsf_internal): Ditto.
101 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
102 various splats with the proper size immediate. Reorder the various
103 cases by ascending size of immediate, and put all such together.
105 2018-09-25 Richard Biener <rguenther@suse.de>
108 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
110 (maybe_create_die_with_external_ref): Do not create
111 DW_TAG_imported_unit here.
112 (add_abstract_origin_attribute): Handle external BLOCK refs.
113 (dwarf2out_abstract_function): Simplify LTO case.
114 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
115 rather than using maybe_create_die_with_external_ref.
117 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
119 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
120 (frndintxf2_mask_pm_i387): Ditto.
121 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
122 Enable for !flag_trapping_math.
123 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
124 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
125 Change operand 1 predicate to nonimmediate_operand.
126 (attr "i387_cw"): Remove mask_pm.
127 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
128 (enum ix86_entity): Remove I387_MASK_PM.
129 * config/i386/i386.c (ix86_i387_mode_needed): Do not
131 (ix86_mode_needed): Ditto.
132 (ix86_mode_after): Ditto.
133 (ix86_mode_entry): Ditto.
134 (ix86_mode_exit): Ditto.
135 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
137 2018-09-25 Jakub Jelinek <jakub@redhat.com>
139 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
140 to_update_switch_stmts to vNULL instead of calling create on them
143 2018-09-25 Richard Biener <rguenther@suse.de>
145 PR tree-optimization/87402
146 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
147 (visit_phi): Re-instantiate handling of supposed to be VARYING
148 but non-VARYING backedge value.
150 2018-09-25 Richard Biener <rguenther@suse.de>
153 * dwarf2out.c (struct sym_off_pair): New.
154 (external_die_map): New global.
155 (lookup_decl_die): When in LTO create DIEs lazily from the
157 (lookup_block_die): New function, create DIEs lazily in LTO.
158 (equate_block_to_die): New function.
159 (dwarf2out_die_ref_for_decl): During WPA get the association
160 from the external DIE map.
161 (dwarf2out_register_external_die): Record mapping into the
163 (maybe_create_die_with_external_ref): New function split out from
164 DIE generation part of old dwarf2out_register_external_die.
165 (add_abstract_origin_attribute): Do not return the DIE. When
166 in LTO reference externals directly.
167 (dwarf2out_abstract_function): When in LTO ignore calls for
168 decls with external DIEs (already present abstract instances).
169 (gen_call_site_die): Adjust.
170 (add_high_low_attributes): Likewise.
171 (gen_lexical_block_die): Likewise.
172 (gen_inlined_subroutine_die): Likewie.
173 (gen_block_die): Likewise.
174 (dwarf2out_inline_entry): Likewise.
175 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
178 2018-09-25 Martin Liska <mliska@suse.cz>
180 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
181 integers and not by a float value.
183 2018-09-25 Martin Liska <mliska@suse.cz>
186 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
188 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
189 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
190 * godump.c (go_format_type): Remove extra parenthesis.
192 2018-09-25 Martin Liska <mliska@suse.cz>
194 * alias.c (set_dest_equal_p): Remove unused function.
195 * config/i386/i386.c (def_builtin_pure2): Likewise.
196 * diagnostic-show-locus.c (class layout): Remove
198 (layout::layout): Likewise here.
199 * dump-context.h (class temp_dump_context): Likewise.
200 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
201 (add_AT_loclistsptr): Likewise.
202 (add_AT_offset): Likewise.
203 (get_AT_hi_pc): Likewise.
204 (is_comdat_die): Likewise.
205 (type_is_enum): Likewise.
207 (add_AT_vms_delta): Likewise.
208 (is_class_die): Likewise.
209 * edit-context.c (class line_event): Remove unused field.
210 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
212 * ipa-cp.c (ipa_get_vr_lat): Likewise.
213 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
214 (ok_for_base_p_nonstrict): Likewise.
215 * tree-chrec.c (is_not_constant_evolution): Likewise.
216 (chrec_fold_poly_cst): Likewise.
217 * tree-if-conv.c (has_pred_critical_p): Likewise.
218 * tree-ssa-coalesce.c (print_exprs): Likewise.
219 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
220 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
221 * tree-vrp.c (value_ranges_intersect_p): Likewise.
222 (value_range_nonnegative_p): Likewise.
224 2018-09-25 Martin Liska <mliska@suse.cz>
226 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
227 Do not handle "GNU Pascal".
228 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
230 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
231 from documentation. Likewise.
232 * dbxout.c (dbxout_range_type): Likewise.
233 * doc/cpp.texi: Likewise.
234 * doc/extend.texi: Likewise.
235 * doc/frontends.texi: Likewise.
236 * doc/invoke.texi: Remove Pascal entry.
237 * tree.def (CLEANUP_POINT_EXPR): Likewise.
238 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
240 2018-09-25 Martin Liska <mliska@suse.cz>
243 * doc/invoke.texi: Document all parameters and remove default
246 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
249 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
250 contains HOST_WIDE_INTs when computing its size.
252 2018-09-24 Jim Wilson <jimw@sifive.com>
255 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
257 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
259 2018-09-24 Andrew Pinski <apinski@marvell.com>
261 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
262 access prev before checking it for NULLness in the
263 AARCH64_FUSE_CMP_BRANCH case.
265 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
268 * config/i386/i386.c (rest_of_insert_endbranch): Set
269 endbr_queued_at_entrance to true and don't insert ENDBR if
270 x86_function_profiler will be called.
271 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
273 * config/i386/i386.h (machine_function): Add
274 endbr_queued_at_entrance.
276 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
278 * genattrtab.c (mk_attr_alt): Use alternative_mask.
279 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
281 (check_attr_test): Use alternative_mask.
282 (get_attr_value): Likewise.
283 (compute_alternative_mask): Use alternative_mask and XWINT.
284 (make_alternative_compare): Use alternative_mask.
285 (attr_alt_subset_p): Use XWINT.
286 (attr_alt_subset_of_compl_p): Likewise.
287 (attr_alt_intersection): Use alternative_mask and XWINT.
288 (attr_alt_union): Likewise.
289 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
290 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
291 (simplify_test_exp): Use alternative_mask and XWINT.
292 (write_test_expr): Use alternative_mask and XWINT, adjust bit
293 number calculation to support 64 bits. Generate code that
295 (main): Use alternative_mask.
296 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
298 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
301 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
302 RETURN+USE when returning via %r14.
304 2018-09-24 Martin Liska <mliska@suse.cz>
306 * gcov.c (output_lines): Print colorization legend
307 for both flag_use_colors and flag_use_hotness_colors.
310 2018-09-24 Martin Liska <mliska@suse.cz>
312 * coverage.c (get_coverage_counts): Use warning_at
313 with current_function_decl location. Use %qD in warning
316 2018-09-24 Martin Liska <mliska@suse.cz>
318 * memory-block.h (memory_block_pool::release): Annotate with
319 valgrind that the memory is not accessible.
321 2018-09-24 Martin Liska <mliska@suse.cz>
324 * asan.c: Make asan_handled_variables extern.
326 * cfgexpand.c (expand_stack_vars): Make sure
327 a representative is unpoison if another
328 variable in the partition is handled by
329 use-after-scope sanitization.
331 2018-09-24 Richard Biener <rguenther@suse.de>
333 PR tree-optimization/63155
334 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
335 the worklist when the edge of the respective argument isn't
338 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
340 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
341 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
342 (MASK_CLASS_P): Update for rename.
343 (MAYBE_MASK_CLASS_P): Ditto.
344 (REG_CLASS_NAMES): Update.
345 (REG_CLASS_CONTENT): Update.
346 * config/i386/i386.c (regclass_map): Update for MASK_REG
347 and ALL_MASK_REGS rename.
348 * config/i386/constraints.md (Yk): Update for rename.
351 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
353 * config/i386/i386.h (enum reg_class): Remove
354 EVEX_SSE_REGS and MOD4_SSE_REGS.
355 (REG_CLASS_NAMES): Update.
356 (REG_CLASS_CONTENT): Update.
357 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
358 registers as ALL_SSE_REGS.
359 (ix86_additional_allocno_class_p): Remove.
360 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
361 (ix86_register_priority): Lower priority of EVEX SSE registers.
362 Use IN_RANGE macro where appropriate.
363 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
364 AVX-5124VNNIW checks.
365 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
366 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
367 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
368 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
369 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
370 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
371 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
372 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
373 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
374 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
375 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
376 * config/i386/constraints.md (Yh): Remove.
378 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
380 * config/i386/i386.c (regclass_map): Declare integer REX registers
383 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
385 * doc/service.texi (Service): Switch the fsf.org link to https.
387 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
390 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
391 Define to speculation_safe_value_not_needed.
393 2018-09-21 Florian Weimer <fweimer@redhat.com>
396 * doc/extend.texi (Common Function Attributes): Mention that
397 noreturn suppresses tail call optimization.
399 2018-09-21 Jeff Law <law@redhat.com>
401 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
402 vr_values::cleanup_edges_and_switches.
403 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
405 (identify_jump_threads): Remove EDGE_IGNORE handling.
406 (execute_vrp): Move handling of to_remove_edges and
407 to_update_switch_stmts into vr_values class member functions.
408 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
409 (to_update_switch_stmts): Likewise.
410 * vr-values.c: Include cfghooks.h.
411 (vr_values::vr_values): Initialize to_remove_edges and
412 to_update_switch_stmts.
413 (vr_values::~vr_values): Verify to_remove_edges and
414 to_update_switch_stmts are empty.
415 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
416 (vr_values::cleanup_edges_and_switches): New member function.
417 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
418 function. Add new data members.
420 2018-09-21 David Malcolm <dmalcolm@redhat.com>
422 PR tree-optimization/87309
423 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
424 calls with pflags and alt_flags.
425 (selftest::test_capture_of_dump_calls): Add test of interaction of
426 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
428 2018-09-21 Olivier Hainque <hainque@adacore.com>
430 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
432 2018-09-21 Olivier Hainque <hainque@adacore.com>
434 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
436 2018-09-21 Olivier Hainque <hainque@adacore.com>
438 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
439 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
440 (VXWORKS_LIBS_RTP): Minor reordering.
442 2018-09-21 Olivier Hainque <hainque@adacore.com>
444 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
445 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
447 2018-09-21 Olivier Hainque <hainque@adacore.com>
449 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
450 (PTRDIFF_TYPE): Likewise.
452 2018-09-21 Olivier Hainque <hainque@adacore.com>
454 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
455 triplet, similar to support for VxWorks7.
456 * config/vxworks-dummy.h: Provide a default definition
457 of TARGET_VXWORKS64 to 0.
459 2018-09-21 Olivier Hainque <hainque@adacore.com>
461 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
462 * config/vxworks-dummy.h: here.
464 2018-09-21 Olivier Hainque <hainque@adacore.com>
466 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
468 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
469 Bo Zhou <zbo.zhou@hisilicon.com>
471 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
472 * config/aarch64/aarch64-tune.md: Regenerated.
473 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
474 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
475 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
477 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
478 Julian Brown <julian@codesourcery.com>
480 * builtins.c (get_builtin_sync_mem): Handle address spaces.
482 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
484 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
485 if the call takes a static chain.
487 2018-09-21 Martin Liska <mliska@suse.cz>
489 * auto-profile.c (autofdo_source_profile::read): Do not
491 (read_profile): Do not add working sets.
492 (read_autofdo_file): Remove sum_all.
493 (afdo_callsite_hot_enough_for_early_inline): Remove const
495 * coverage.c (struct counts_entry): Remove gcov_summary.
496 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
497 do not support GCOV_TAG_PROGRAM_SUMMARY.
498 (get_coverage_counts): Remove summary and expected
500 * coverage.h (get_coverage_counts): Likewise.
501 * doc/gcov-dump.texi: Remove -w option.
502 * gcov-dump.c (dump_working_sets): Remove.
503 (main): Do not support '-w' option.
504 (print_usage): Likewise.
505 (tag_summary): Likewise.
506 * gcov-io.c (gcov_write_summary): Do not dump
508 (gcov_read_summary): Likewise.
509 (gcov_histo_index): Remove.
510 (gcov_histogram_merge): Likewise.
511 (compute_working_sets): Likewise.
512 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
514 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
515 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
516 (GCOV_HISTOGRAM_SIZE): Remove.
517 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
518 (struct gcov_summary): Simplify rapidly just
519 to runs and sum_max fields.
520 (gcov_histo_index): Remove.
521 (NUM_GCOV_WORKING_SETS): Likewise.
522 (compute_working_sets): Likewise.
523 * gcov-tool.c (print_overlap_usage_message): Remove
525 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
526 (output_lines): Remove program related line.
527 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
528 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
530 (input_profile_summary): Do not read it.
531 (merge_profile_summaries): And do not merge it.
532 (input_symtab): Do not call removed function.
533 * modulo-sched.c (sms_schedule): Do not print sum_max.
534 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
535 removed when histogram method was invented.
536 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
538 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
540 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
541 and dump selected value.
542 * profile.c (add_working_set): Remove.
543 (get_working_sets): Likewise.
544 (find_working_set): Likewise.
545 (get_exec_counts): Do not work with working sets.
546 (read_profile_edge_counts): Do not inform as sum_max is removed.
547 (compute_branch_probabilities): Likewise.
548 (compute_value_histograms): Remove argument for call of
550 * profile.h: Do not make gcov_summary const.
552 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
554 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
556 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
558 PR tree-optimization/86990
559 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
560 Check that the entire merged store group is made of constants only for
563 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
565 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
566 (VTABLE_VERIFICATION_SPEC): Likewise.
567 (SANITIZER_EARLY_SPEC): Likewise.
568 (SANITIZER_SPEC): Likewise.
569 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
570 * doc/invoke.texi (Link Options): Document -r.
572 2018-09-20 Richard Biener <rguenther@suse.de>
575 * gimplify.c (gimplify_expr): Retain alignment of
576 addressable lvalue in dereference.
578 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
581 * configure.ac: Check for .loc is_stmt support.
582 * configure, config.in: Rebuilt.
583 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
586 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
588 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
589 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
592 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
594 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
596 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
599 2018-09-20 Marek Polacek <polacek@redhat.com>
601 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
603 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
605 PR tree-optimization/87288
606 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
607 into account when determining PEELING_FOR_NITERS.
609 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
611 PR tree-optimization/86877
612 * tree-vect-loop.c (vect_analyze_loop_2): Call
613 vect_verify_datarefs_alignment.
615 2018-09-19 Marek Polacek <polacek@redhat.com>
617 * doc/invoke.texi: Document -Wclass-conversion.
619 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
621 * config/pa/pa.c (pa_adjust_priority): Delete.
622 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
624 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
625 (atomic_storehi): Likewise.
626 (atomic_storesi): Likewise.
627 (atomic_loaddi): Restore compare and swap exchange loop code.
629 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
631 PR rtl-optimization/86902
632 * combine.c (try_combine): When changing the CC mode used, don't change
633 an unrelated mode in other_insn to that new CC mode.
635 2018-09-19 David Malcolm <dmalcolm@redhat.com>
637 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
638 with %T in place of calls to dump_generic_expr.
639 (prune_runtime_alias_test_list): Likewise.
640 (create_runtime_alias_checks): Likewise.
641 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
642 (vect_analyze_data_ref_dependence): Likewise.
643 (vect_slp_analyze_data_ref_dependence): Likewise.
644 (vect_record_base_alignment): Likewise. Use %G in place of call
646 (vect_compute_data_ref_alignment): Likewise.
647 (verify_data_ref_alignment): Likewise.
648 (vect_find_same_alignment_drs): Likewise.
649 (vect_analyze_group_access_1): Likewise.
650 (vect_analyze_data_ref_accesses): Likewise.
651 (dependence_distance_ge_vf): Likewise.
652 (dump_lower_bound): Likewise.
653 (vect_prune_runtime_alias_test_list): Likewise.
654 (vect_find_stmt_data_reference): Likewise.
655 (vect_analyze_data_refs): Likewise.
656 (vect_create_addr_base_for_vector_ref): Likewise.
657 (vect_create_data_ref_ptr): Likewise.
658 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
659 (vect_can_advance_ivs_p): Likewise.
660 (vect_update_ivs_after_vectorizer): Likewise.
661 (vect_gen_prolog_loop_niters): Likewise.
662 (vect_prepare_for_masked_peels): Likewise.
663 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
664 (vect_determine_vectorization_factor): Likewise.
665 (vect_is_simple_iv_evolution): Likewise.
666 (vect_analyze_scalar_cycles_1): Likewise.
667 (vect_analyze_loop_operations): Likewise.
668 (report_vect_op): Likewise.
669 (vect_is_slp_reduction): Likewise.
670 (check_reduction_path): Likewise.
671 (vect_is_simple_reduction): Likewise.
672 (vect_create_epilog_for_reduction): Likewise.
673 (vect_finalize_reduction:): Likewise.
674 (vectorizable_induction): Likewise.
675 (vect_transform_loop_stmt): Likewise.
676 (vect_transform_loop): Likewise.
677 (optimize_mask_stores): Likewise.
678 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
679 (vect_split_statement): Likewise.
680 (vect_recog_over_widening_pattern): Likewise.
681 (vect_recog_average_pattern): Likewise.
682 (vect_determine_min_output_precision_1): Likewise.
683 (vect_determine_precisions_from_range): Likewise.
684 (vect_determine_precisions_from_users): Likewise.
685 (vect_mark_pattern_stmts): Likewise.
686 (vect_pattern_recog_1): Likewise.
687 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
688 (vect_record_max_nunits): Likewise.
689 (vect_build_slp_tree_1): Likewise.
690 (vect_build_slp_tree_2): Likewise.
691 (vect_print_slp_tree): Likewise.
692 (vect_analyze_slp_instance): Likewise.
693 (vect_detect_hybrid_slp_stmts): Likewise.
694 (vect_detect_hybrid_slp_1): Likewise.
695 (vect_slp_analyze_operations): Likewise.
696 (vect_slp_analyze_bb_1): Likewise.
697 (vect_transform_slp_perm_load): Likewise.
698 (vect_schedule_slp_instance): Likewise.
699 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
700 (vect_mark_stmts_to_be_vectorized): Likewise.
701 (vect_init_vector_1): Likewise.
702 (vect_get_vec_def_for_operand): Likewise.
703 (vect_finish_stmt_generation_1): Likewise.
704 (vect_check_load_store_mask): Likewise.
705 (vectorizable_call): Likewise.
706 (vectorizable_conversion): Likewise.
707 (vectorizable_operation): Likewise.
708 (vectorizable_load): Likewise.
709 (vect_analyze_stmt): Likewise.
710 (vect_is_simple_use): Likewise.
711 (vect_get_vector_types_for_stmt): Likewise.
712 (vect_get_mask_type_for_stmt): Likewise.
713 * tree-vectorizer.c (increase_alignment): Likewise.
715 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
717 * doc/rtl.texi: Adjust vec_select description.
718 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
719 non-constant selectors.
721 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
723 * config/aarch64/aarch64-protos.h
724 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
725 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
726 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
727 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
728 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
729 (AARCH64_FL_PROFILE): Move index so flags are ordered.
730 (AARCH64_ISA_RCPC8_4): New flag.
731 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
732 to aarch64_offset_9bit_signed_unscaled_p.
733 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
735 * config/aarch64/constraints.md (Ust): New constraint.
736 * config/aarch64/predicates.md.
737 (aarch64_9bit_offset_memory_operand): New predicate.
738 (aarch64_rcpc_memory_operand): New predicate.
740 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
742 PR rtl-optimization/87361
743 * rtlanal.c (nonzero_bits1): Revert accidental change.
745 2018-09-19 Richard Biener <rguenther@suse.de>
747 PR tree-optimization/87349
748 PR tree-optimization/87342
749 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
751 2018-09-18 Marek Polacek <polacek@redhat.com>
753 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
754 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
756 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
758 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
760 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
762 PR rtl-optimization/86882
763 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
765 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
767 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
768 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
770 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
773 * doc/invoke.texi (Link Options): Fix formatting and grammar.
775 2018-09-18 Richard Biener <rguenther@suse.de>
778 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
779 (compute_samebase_partition_bases): Likewise.
780 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
781 (gimple_can_coalesce_p): Simplify.
783 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
785 Handle a library implementation of ffs calling __builtin_ffs.
786 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
787 (mmix_init_libfuncs): New function: make __builtin_ffs expand
790 2018-09-17 David Malcolm <dmalcolm@redhat.com>
792 * diagnostic-show-locus.c (class layout_range): Add field
794 (layout_range::layout_range): Add "original_idx" param and use it
795 to initialize new field.
796 (make_range): Use 0 for original_idx.
797 (layout::layout): Pass in index to calls to
798 maybe_add_location_range.
799 (layout::maybe_add_location_range): Add param "original_idx" and
800 pass it on to layout_range.
801 (layout::print_any_labels): Pass on range->m_original_idx to
803 (gcc_rich_location::add_location_if_nearby): Use 0 for
805 * gcc-rich-location.h (text_range_label::get_text): Update for new
807 (range_label_for_type_mismatch::get_text): Likewise.
809 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
811 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
813 2018-09-17 David Malcolm <dmalcolm@redhat.com>
815 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
816 format_string_diagnostic_t.
817 (fmtwarn_n): Likewise.
818 * substring-locations.c
819 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
820 (format_warning_n_va): Convert to...
821 (format_string_diagnostic_t::emit_warning_n_va): ...this.
822 (format_warning_va): Convert to...
823 (format_string_diagnostic_t::emit_warning_va): ...this.
824 (format_warning_at_substring): Convert to...
825 (format_string_diagnostic_t::emit_warning): ...this.
826 (format_warning_at_substring_n): Convert to...
827 (format_string_diagnostic_t::emit_warning_n): ...this.
828 * substring-locations.h (class format_string_diagnostic_t): New
830 (format_warning_va): Convert to
831 format_string_diagnostic_t::emit_warning_va.
832 (format_warning_n_va): Convert to
833 format_string_diagnostic_t::emit_warning_n_va.
834 (format_warning_at_substring): Convert to
835 format_string_diagnostic_t::emit_warning.
836 (format_warning_at_substring_n): Convert to
837 format_string_diagnostic_t::emit_warning_n.
839 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
840 Bernd Schmidt <bernds_cb1@t-online.de>
842 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
845 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
847 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
848 operand 0 predicate to nonimmediate operand.
849 (rint<mode>2_frndint): Remove insn pattern.
850 (rint<mode>2): Change operand 1 predicate to general_operand.
851 Extend operand 1 to XFmode and generate rintxf2 insn.
852 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
853 Do not use X87MODEF mode macro.
854 (frndintxf2_<rounding>_i387): Rename from
855 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
856 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
857 to XFmode and generate significandxf3 insn.
859 2018-09-17 Richard Biener <rguenther@suse.de>
861 PR tree-optimization/87328
862 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
863 visiting unexecutable backedges when not iterating.
864 (do_rpo_vn): Mark all edges not executable even when not
867 2018-09-17 Martin Jambor <mjambor@suse.cz>
870 * doc/invoke.texi (Warning Options): Likewise.
872 2018-09-17 Richard Biener <rguenther@suse.de>
874 PR tree-optimization/87301
875 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
876 clean EH info from leftover copy assignments.
878 2018-09-17 Martin Liska <mliska@suse.cz>
880 PR gcov-profile/85871
881 * gcov.c (output_intermediate_file): Fix out of bounds
884 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
886 * config/arc/arc.c: Object attributes for core4 not reflected
888 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
891 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
893 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
895 2018-09-17 Martin Liska <mliska@suse.cz>
897 * doc/gcov.texi: Document new option --use-hotness-colors.
898 * gcov.c (struct source_info): Declare new field.
899 (source_info::source_info): Set default for maximum_count.
900 (print_usage): Add new -q option.
901 (process_args): Process it.
902 (accumulate_line_info): Save src->maximum_count.
903 (output_line_beginning): Make color line number if
904 flag_use_hotness_colors is set.
905 (output_line_details): Pass default argument value.
906 (output_lines): Pass src->maximum_count.
908 2018-09-17 Martin Liska <mliska@suse.cz>
910 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
911 Use processor_names table.
912 * config/i386/i386.c (ix86_default_align): Use
913 processor_cost_table for alignment values.
914 (ix86_option_override_internal): Use processor_names.
915 (ix86_function_specific_print): Likewise.
916 * config/i386/i386.h (struct processor_costs):
917 Add alignment values.
918 (struct ptt): Remove and replace with const char *.
919 * config/i386/x86-tune-costs.h (struct processor_costs):
920 Declare default alignments for all costs.
922 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
924 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
925 symbolics or VR_VARYING ranges for ABS_EXPR.
926 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
927 when range will wrap.
929 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
932 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
933 before and after a JUMP_TABLE_DATA.
935 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
938 * dojump.c (do_compare_and_jump): Canonicalize function pointers
939 when one operand is a function pointer. Use POINTER_TYPE_P and
940 FUNC_OR_METHOD_TYPE_P.
941 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
942 * fold-const.c (build_range_check): Likewise.
943 * match.pd (simple_comparison): Likewise.
945 2018-09-14 David Malcolm <dmalcolm@redhat.com>
948 * spellcheck.c (get_edit_distance_cutoff): New function.
949 (selftest::test_edit_distance_unit_test_oneway): Rename to...
950 (selftest::test_get_edit_distance_one_way): ...this.
951 (selftest::test_get_edit_distance_unit): Rename to...
952 (selftest::test_get_edit_distance_both_ways): ...this.
953 (selftest::test_edit_distances): Move tests to this new function,
954 and test some more pairs of strings. Update for above renaming.
955 (selftest::get_old_cutoff): New function.
956 (selftest::test_get_edit_distance_cutoff): New function.
957 (selftest::assert_suggested_for): New function.
958 (ASSERT_SUGGESTED_FOR): New macro.
959 (selftest::assert_not_suggested_for): New function.
960 (ASSERT_NOT_SUGGESTED_FOR): New macro.
961 (selftest::test_suggestions): New function.
962 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
963 tests to selftest::test_edit_distances and call it. Add calls to
964 selftest::test_get_edit_distance_cutoff and
965 selftest::test_suggestions.
966 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
967 (best_match::consider): Replace hard-coded cutoff calculation with
969 (best_match::get_cutoff): New declaration.
970 (best_match::get_best_meaningful_candidate): Likewise.
972 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
974 * builtins.c (fold_builtin_strlen): Remove TODO comment.
976 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
979 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
981 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
982 terminated string literal.
984 2018-09-14 Martin Sebor <msebor@redhat.com>
986 * builtins.c (unterminated_array): Handle ARRAY_REF.
987 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
988 * builtins.h (unterminated_array): Declare extern.
989 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
991 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
994 2018-09-14 Martin Sebor <msebor@redhat.com>
995 Jeff Law <law@redhat.com>
997 * builtins.c (unterminated_array): New.
998 (expand_builtin_strcpy): Adjust.
999 (expand_builtin_strcpy_args): Detect unterminated arrays.
1000 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
1001 unterminated arrays.
1002 * gimple-fold.h (get_maxval_strlen): Add argument.
1003 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
1005 * gimple-fold.c (get_range_strlen): Add argument.
1006 (get_maxval_strlen): Adjust.
1007 * gimple-fold.h (get_range_strlen): Add argument.
1009 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
1011 * config/i386/movdirintrin.h: Fix copyright year.
1013 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
1015 * reg-stack.c: Include regs.h.
1016 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
1017 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
1018 FIRST_STACK_REG, not DFmode.
1019 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
1020 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
1021 (change stack): Default register mode to the reg_raw_mode of
1022 FIRST_STACK_REG, not DFmode.
1023 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
1024 (*swapxf): Rename from swapxf.
1026 2018-09-14 Carl Love <cel@us.ibm.com>
1028 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
1029 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
1031 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
1034 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
1037 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
1040 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
1042 2018-09-14 Jason Merrill <jason@redhat.com>
1044 Fix --enable-gather-detailed-mem-stats.
1045 * hash-table.c (hash_table_usage): Change from variable to function.
1046 * hash-table.h: Adjust.
1047 * Makefile.in: Add missing dependencies on hash-table.h.
1049 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1051 PR tree-optimization/87259
1053 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
1054 execute_cse_reciprocals_1 has tried transforming.
1056 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
1058 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
1059 VR_VARYING for PLUS/MINUS_EXPR.
1061 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
1063 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
1066 2018-09-14 Richard Biener <rguenther@suse.de>
1069 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
1070 bits for the merged partition.
1072 2018-09-13 Martin Sebor <msebor@redhat.com>
1073 Bernd Edlinger <bernd.edlinger@hotmail.de>
1075 * builtins.h (c_srlen): Add argument.
1076 * builtins.c (warn_string_no_nul): New function.
1077 (c_strlen): Add argument and use it. Update recursive calls.
1078 Pass DECL argument to string_constant to get info on non
1079 terminated strings. Update *NONSTR as needed.
1080 (fold_builtin_strlen): Add argument to calls to c_strlen.
1081 Warn for unterminated arrays.
1082 (warn_string_no_null): Add prototype.
1083 * expr.c (string_constant): Update arguments. Update recursive
1084 calls appropriately. Detect missing NUL terminator and outermost
1085 declaration its missing in.
1086 Improve checks for arrays with nonzero lower bound or elements
1087 that are not a single byte. Simplify offset computation.
1088 Simplify checks for non-NUL terminated strings.
1089 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
1090 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
1092 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
1094 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
1096 * fold-const.c (c_getstr): Fix function comment. Remove unused third
1097 argument. Fix range checks.
1098 * fold-const.h (c_getstr): Adjust protoype.
1099 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
1100 string is constant but contains no NUL byte.
1102 * expr.c (string_constant): Adjust function comment.
1103 Remove bogus check for zero termination.
1105 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
1107 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
1108 (get_constant_size): Don't make STRING_CSTs larger than they are.
1109 (check_string_literal): New check function for STRING_CSTs.
1110 (output_constant): Use it.
1112 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
1115 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
1117 2018-09-13 Richard Biener <rguenther@suse.de>
1119 PR tree-optimization/87263
1120 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
1121 (struct unwind_state): Add max_rpo field.
1122 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
1123 Compute max_rpo, the max RPO number a block can be backwards reached
1124 from. Re-write non-iterating mode to a RPO ordered worklist approach,
1125 separating it from the iterating mode.
1127 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1129 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
1130 (rfs_decision): New scheduling decision.
1132 2018-09-13 Richard Biener <rguenther@suse.de>
1135 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
1136 (vn_nary_op_insert_pieces_predicated): Do not write useless
1137 valid_dominated_by_p entry outside of the allocated storage.
1139 2018-09-13 Omar Sandoval <osandov@osandov.com>
1140 Tom de Vries <tdevries@suse.de>
1143 * dwarf2out.c (is_c): New function.
1144 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
1146 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
1149 * config/aarch64/aarch64.md (*aarch64_bfxil):
1151 * config/aarch64/constraints.md (Ulc): Define.
1152 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
1154 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
1157 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1159 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
1160 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
1161 aarch64_layout_frame call.
1162 (aarch64_expand_epilogue): Likewise.
1163 (aarch64_initial_elimination_offset): Likewise.
1164 (aarch64_get_separate_components): Likewise.
1165 (aarch64_use_return_insn_p): Likewise.
1166 (aarch64_layout_frame): Remove unneeded check.
1168 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1170 * configure.ac: Only append
1171 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
1172 gcc_config_arguments if it was never reconfigured or last reconfigure
1173 was with different arguments.
1174 * configure: Regenerated.
1176 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1177 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1180 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
1181 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
1183 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1185 PR tree-optimization/87287
1186 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
1187 X % C == 0 to X % (unsigned) C == 0 optimization to ...
1188 * match.pd (X % C == 0): ... here. New optimization.
1190 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1193 * expr.h (maybe_optimize_mod_cmp): Declare.
1194 * expr.c (mod_inv): New function.
1195 (maybe_optimize_mod_cmp): New function.
1196 (do_store_flag): Use it.
1197 * cfgexpand.c (expand_gimple_cond): Likewise.
1199 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
1200 Julian Brown <julian@codesourcery.com>
1203 * gimplify.c (gimplify_scan_omp_clauses): Set
1204 target_firstprivatize_array_bases in OpenACC parallel and kernels
1205 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
1206 OpenACC data regions.
1208 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
1210 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
1211 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
1213 2018-09-12 Richard Biener <rguenther@suse.de>
1215 PR tree-optimization/87280
1216 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
1217 edge but unreachable target.
1218 (do_rpo_vn): For conservatively handling a PHI only mark
1219 the backedge executable but not the block reachable.
1221 2018-09-12 Richard Biener <rguenther@suse.de>
1223 PR tree-optimization/87266
1224 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
1227 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
1229 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
1231 ("trunc<BFP:mode><DFP_ALL:mode>2")
1232 ("trunc<DFP_ALL:mode><BFP:mode>2")
1233 ("extend<BFP:mode><DFP_ALL:mode>2")
1234 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
1235 according to the target operand type.
1237 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1238 Andreas Krebbel <krebbel@linux.ibm.com>
1240 PR tree-optimization/86844
1241 * gimple-ssa-store-merging.c
1242 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
1243 there are any overlapping stores in between them, make sure they are
1244 also coalesced or we give up completely.
1246 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1249 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
1250 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
1252 2018-09-12 Tom de Vries <tdevries@suse.de>
1254 * common.opt (gdescribe-dies): Add option.
1255 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
1256 attribute for artifical and nameless decls.
1257 (dwarf2out_register_external_die): Add description attribute to
1258 external reference die.
1259 (add_desc_attribute): New functions.
1260 (gen_subprogram_die): Add description attribute to
1261 DW_TAG_call_site_parameter.
1262 * tree-pretty-print.c (print_generic_expr_to_str): New function.
1263 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
1264 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
1266 (@item -gdescribe-dies): Add.
1268 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1270 * tree-vrp.c (vrp_shift_undefined_p): Remove.
1271 (extract_range_from_binary_expr_1: Call
1272 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
1273 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
1276 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1278 * gimple-ssa-warn-alloca.c
1279 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
1280 field for ALLOCA_BOUND_*_LARGE.
1282 2018-09-11 Nathan Sidwell <nathan@acm.org>
1284 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
1286 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1288 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
1289 for clobbers. Remove obsolete comment.
1291 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1293 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
1294 mpxchk, mpxld and mpxst types.
1295 (define_attr length_immediate): Remove all processing of mpx types.
1296 (define_attr prefix_0f): Ditto.
1297 (define_attr memory): Ditto.
1299 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1301 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
1302 (log<mode>2): Change operand 1 predicate to general_operand.
1303 Extend operand 1 to XFmode and generate logxf3 insn.
1304 (log10<mode>2): Change operand 1 predicate to general_operand.
1305 Extend operand 1 to XFmode and generate log10xf3 insn.
1306 (log2<mode>2): Change operand 1 predicate to general_operand.
1307 Extend operand 1 to XFmode and generate log2xf3 insn.
1308 (fyl2xp1_extend<mode>xf3_i387): Remove.
1309 (log1p<mode>2): Change operand 1 predicate to general_operand.
1310 Extend operand 1 to XFmode and generate log1pxf3 insn.
1311 (fxtract_extend<mode>xf3_i387): Remove.
1312 (logb<mode>2): Change operand 1 predicate to general_operand.
1313 Extend operand 1 to XFmode and generate logbxf3 insn.
1314 (ilogb<mode>2): Change operand 1 predicate to general_operand.
1315 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
1316 (significand<mode>2): Change operand 1 predicate to general_operand.
1317 Extend operand 1 to XFmode and generate significandxf3 insn.
1319 2018-09-11 Nathan Sidwell <nathan@acm.org>
1321 * gcc.c (perror_with_name, pfatal_with_name): Delete.
1322 (load_specs): Use fatal_error.
1323 (DELETE_IF_ORDINARY, process_command): Use error.
1324 (execute, run_attempt): Use fatal_error.
1326 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
1328 * diagnostic-core.h (sorry_at): New prototype.
1329 * diagnostic.c (sorry_at): New function.
1331 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
1333 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
1334 by zero as VR_UNDEFINED.
1336 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1338 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
1339 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
1340 (<sincos>mode2): New expander.
1341 (sincos_extend<mode>xf3_i387): Remove insn pattern.
1342 (sincos -> sin, cos splitters): Remove splitter patterns.
1343 (sincos<mode>3): Change operand 2 predicate to general_operand.
1344 Extend operand 2 to XFmode and generate sincosxf3 insn.
1345 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
1346 Change operand 3 predicate to const1_operand.
1347 (fptan_extend<mode>xf4_i387): Remove insn pattern.
1348 (tanxf2): Update operands in the call to fptanxf4_i387.
1349 (tan<mode>2): Change operand 1 predicate to general_operand.
1350 Extend operand 1 to XFmode and generate tanxf3 insn.
1351 (atan2xf3): Rename from *fpatanxf3_i387.
1352 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
1353 (atan2xf3): Remove expander.
1354 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
1355 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
1356 (atan<mode>2): Change operand 1 predicate to general_operand.
1357 Extend operand 1 to XFmode and generate atanxf3 insn.
1359 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1361 * config/i386/i386.md (x87/SSE constant load splitter): Use
1362 memory_operand instead of nonimmediate_operand for input operand
1365 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
1367 * config/i386/i386.md (float partial SSE register stall splitter): Move
1368 splitter near its instruction pattern.
1369 (float_extend partial SSE register stall splitter): Ditto.
1370 (float_truncate partial SSE register stall splitter): Ditto.
1372 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
1375 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1376 to speculation_safe_value_not_needed.
1379 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
1380 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
1381 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
1382 leaf_function_p, instead use has_hard_reg_initial_val.
1384 2018-09-09 Nathan Sidwell <nathan@acm.org>
1386 * gcc.h (pfatal_with_name): Don't declare here.
1387 * gcc.c (pfatal_with_name): Make static.
1389 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
1391 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
1394 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
1396 PR rtl-optimization/85458
1397 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
1398 priority hook to reduce the priority of EXPR.
1400 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
1402 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
1403 DImode for x87 on 32bit targets. Conditionally disable x87 modes
1404 with X87_ENABLE_FLOAT. Remove preparation code.
1405 (*float<SWI48:mode><MODEF:mode>2): Rename from
1406 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
1407 math using "enabled" attribute.
1408 (*floatdi<MODEF:mode>2_i387): Rename from
1409 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
1410 enable for 32bit targets only.
1411 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
1413 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
1414 as operand 1 predicate. Rewrite as define_insn_and_split.
1415 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
1417 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
1419 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
1420 to fallthru to FLOAT case.
1422 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1425 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
1426 around folding of vec_sl to handle out of range shift values.
1428 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
1430 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
1431 Update callers to gen_fix_trunc<mode>_i387_fisttp
1432 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
1433 nonimmediate_operand.
1434 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
1435 and corresponding splitters.
1436 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
1437 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
1438 (fix_truncdi_i387_with_temp): Remove insn pattern
1439 and corresponding splitters.
1440 (fix_trunc<mode>_i387): Change operand 0 predicate to
1441 nonimmediate_operand.
1442 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1443 and corresponding splitters.
1444 (*fistdi2_1): Remove.
1446 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1447 (lrintxfdi2): Remove expander. Reimplement as define_insn.
1448 (*fist<mode>2_1): Remove.
1449 (fist<mode>2): Ditto.
1450 (fist<mode>2_with_temp): Remove insn pattern and corresponding
1452 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
1453 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1454 (fistdi2_<rounding>): Change operand 0 predicate to
1455 nonimmediate_operand.
1456 (fistdi2_<rounding>_with_temp): Remove insn pattern
1457 and corresponding splitters.
1458 (fist<mode>2_<rounding>): Change operand 0 predicate to
1459 nonimmediate_operand.
1460 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1461 and corresponding splitters.
1463 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1465 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
1467 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1470 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1472 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1473 early gimple folding of vec_splat().
1474 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1475 * gimple-fold.h: Add an extern define for tree_vec_extract().
1477 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1479 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1480 wrappers around TREE_TYPE comparisons.
1482 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1485 * config/s390/predicates.md: Add nonsym_memory_operand.
1486 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1487 contains a SYMBOL_REF, load it into an intermediate pseudo.
1488 (s390_emit_compare_and_swap): Legitimize operand.
1489 * config/s390/s390.md: Use the new nonsym_memory_operand
1490 with UNSPECV_CAS patterns.
1492 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1495 * config/s390/s390-passes.def: New file.
1496 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1498 (make_pass_s390_early_mach): Add declaration.
1499 * config/s390/s390.c (make_pass_s390_early_mach):
1500 (s390_option_override): Remove dynamic registration.
1501 * config/s390/t-s390: Add s390-passes.def.
1503 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1505 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1506 Remove UNSPEC_LTREL_BASE check.
1507 (annotate_constant_pool_refs): Likewise.
1508 (find_constant_pool_ref): Likewise.
1509 (find_ltrel_base): Removed.
1510 (replace_ltrel_base): Removed.
1511 (s390_mainpool_finish): Remove replace_ltrel_base call.
1512 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1513 (s390_chunkify_finish): Remove replace_ltrel_base call.
1514 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1516 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1519 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1520 to speculation_safe_value_not_needed.
1522 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1523 Bernd Schmidt <bernds_cb1@t-online.de>
1525 * config/nvptx/nvptx-opts.h: New file.
1526 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1527 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1529 (TARGET_SM35): New macro.
1530 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1532 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1534 (misa=): New option.
1535 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1537 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1539 * config/i386/i386.md (truncdfsf2): Remove expander.
1540 (truncdfsf2_with_temp): Ditto.
1541 (truncxf<mode>2): Ditto.
1542 (*truncdfsf_fast_mixed): Remove insn pattern.
1543 (*truncdfsf_fast_i387): Ditto.
1544 (*truncdfsf_mixed): Ditto.
1545 (*truncdfsf_i387): Ditto.
1546 (*truncdfsf2_i387_1): Ditto.
1547 (*truncxfsf2_mixed): Ditto.
1548 (*truncxfdf2_mixed): Ditto.
1549 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1550 to call gen_truncxf<mode>2 instead.
1551 (*truncxf<mode>2_i387): Remove.
1552 (reg->reg splitters): Remove splitter pattern.
1553 (reg->mem splitters): Ditto.
1555 (truncdfsf2): New insn pattern.
1556 (truncxf<mode>2): Ditto.
1558 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1560 * tree-ssa-math-opts.c (is_mult_by): New function.
1561 (is_square_of): Use the above.
1562 (optimize_recip_sqrt): New function.
1563 (pass_cse_reciprocals::execute): Use the above.
1565 2018-09-05 Richard Biener <rguenther@suse.de>
1568 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1569 to zero-init the emplaced vec.
1571 2018-09-05 Martin Liska <mliska@suse.cz>
1573 PR tree-optimization/87205
1574 * tree-switch-conversion.c (pass_lower_switch::execute):
1575 Group cases for switch statements.
1577 2018-09-05 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/87217
1580 * tree-ssa-sccvn.c (vuse_valueize): New.
1581 (vn_reference_lookup_pieces): Use it.
1582 (vn_reference_lookup): Likewise.
1584 2018-09-05 Nathan Sidwell <nathan@acm.org>
1587 * stor-layout.c (place_field): Scan forwards to check last
1588 bitfield when ms_bitfield_placement is in effect.
1590 2018-09-05 Richard Biener <rguenther@suse.de>
1593 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1596 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1597 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1599 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1600 * config.gcc (extra_objs): Build it.
1601 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1603 * config/aarch64/aarch64-passes.def
1604 (pass_tag_collision_avoidance): New pass.
1605 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1606 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1607 (aarch64_classify_address): Remove static qualifier.
1608 (aarch64_address_info, aarch64_address_type): Move to...
1609 * config/aarch64/aarch64-protos.h: ... here.
1610 (make_pass_tag_collision_avoidance): New function.
1611 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1614 2018-09-05 Martin Liska <mliska@suse.cz>
1616 * doc/gcov.texi: Update documentation of humar
1618 * gcov.c (format_count): Print one decimal place, it provides
1619 more fine number of situations like '1G' vs. '1.4G'.
1621 2018-09-05 Martin Liska <mliska@suse.cz>
1624 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1625 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1626 to generate a MASK value.
1628 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1631 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1632 if frame pointer isn't used.
1633 (compute_frame_pointer_to_fb_displacement): Likewise.
1635 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1638 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1639 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1640 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1641 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1642 and OPTION_MASK_ISA_XSAVEC_UNSET.
1644 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1646 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1647 XOR operations in NAND case.
1649 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1651 * wide-int-range.cc (wide_int_range_convert): New.
1652 * wide-int-range.h (wide_int_range_convert): New.
1653 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1654 code into wide_int_range_convert.
1655 (extract_range_into_wide_ints): Do not munge anti range constants
1656 into the entire domain. Just return the range back.
1658 2018-09-04 Martin Liska <mliska@suse.cz>
1660 * genmatch.c (output_line_directive): Add new argument
1662 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1664 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1666 * doc/invoke.texi (Option Summary): Add whitespace.
1668 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1670 2018-09-04 Richard Biener <rguenther@suse.de>
1672 PR tree-optimization/87211
1673 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1674 backedge value we're supposed to treat as VARYING also number
1675 the PHI to VARYING in case it got a different value-number already.
1677 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1679 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1680 (extract_range_from_binary_expr_1): Do not call
1681 vrp_can_optimize_bit_op.
1682 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1684 (wide_int_range_get_mask_and_bounds): New.
1685 (wide_int_range_optimize_bit_op): New.
1686 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1687 (wide_int_range_bit_and): Same.
1688 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1689 (wide_int_range_optimize_bit_op): New.
1690 (wide_int_range_get_mask_and_bounds): New.
1692 2018-09-04 Richard Biener <rguenther@suse.de>
1694 PR tree-optimization/87176
1695 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1696 variable. When value-numbering a virtual PHI node make sure
1697 to not value-number to the backedge value.
1699 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1701 * doc/extend.texi (Long Long, Hex Floats): Document support for
1702 long long and hex floats in more recent versions of ISO C++.
1704 2018-09-03 Richard Biener <rguenther@suse.de>
1706 PR tree-optimization/87177
1707 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1710 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1712 * bb-reorder.c (edge_order): Convert to C-qsort-style
1713 tri-state comparator.
1714 (reorder_basic_blocks_simple): Change std::stable_sort to
1717 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1719 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1720 tri-state comparator.
1721 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1723 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1725 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
1726 (mergesort): ... here as maximum count for using netsort.
1727 (gcc_qsort): Set nlim to 3 if stable sort is requested.
1728 (gcc_stablesort): New.
1729 * system.h (gcc_stablesort): Declare.
1731 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1733 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1734 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
1735 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
1737 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
1739 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1740 lxsdx and stxsdx alternatives.
1741 (*mov<mode>_hardfloat64): Ditto.
1742 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1744 2018-09-03 Richard Biener <rguenther@suse.de>
1746 PR tree-optimization/87200
1747 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1750 2018-09-03 Martin Liska <mliska@suse.cz>
1752 PR tree-optimization/87201
1753 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1754 Fix parenthesis in an expression.
1756 2018-09-03 Richard Biener <rguenther@suse.de>
1758 PR tree-optimization/87197
1759 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1760 visited. CSE the VN_INFO hashtable lookup.
1762 PR tree-optimization/87169
1763 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1764 iterating make sure there's no extra backedges from irreducible
1765 regions feeding the header. Mark the destination block
1768 2018-09-03 Martin Liska <mliska@suse.cz>
1771 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1772 * common/common-targhooks.c (default_get_valid_option_values):
1774 * common/common-targhooks.h (default_get_valid_option_values):
1776 * common/config/i386/i386-common.c: Move processor_target_table
1778 (ix86_get_valid_option_values): New function.
1779 (TARGET_GET_VALID_OPTION_VALUES): New macro.
1780 * config/i386/i386.c (struct ptt): Move to i386-common.c.
1781 (PTA_*): Move all defined masks into i386-common.c.
1782 (ix86_function_specific_restore): Use new processor_cost_table.
1783 * config/i386/i386.h (struct ptt): Moved from i386.c.
1784 (struct pta): Likewise.
1785 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1786 * doc/tm.texi.in: Likewise.
1787 * opt-suggestions.c (option_proposer::suggest_option):
1788 Pass prefix to build_option_suggestions.
1789 (option_proposer::get_completions): Likewise.
1790 (option_proposer::build_option_suggestions): Use the new target
1792 * opts.c (struct option_help_tuple): New struct.
1793 (print_filtered_help): Use the new target hook.
1795 2018-09-03 Martin Liska <mliska@suse.cz>
1798 * predict.c (set_even_probabilities): Add likely_edges
1799 argument and handle cases where we have precisely one
1801 (combine_predictions_for_bb): Catch also likely_edges.
1802 (tree_predict_by_opcode): Handle gswitch statements.
1803 * tree-cfg.h (find_case_label_for_value): New declaration.
1804 (find_taken_edge_switch_expr): Likewise.
1805 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1806 Find pivot in decision tree based on probabily, not by number of
1809 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1811 * doc/standards.texi (Standards): Update Objective-C reference.
1813 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1815 * doc/install.texi (Prerequisites): Update link for MPC.
1817 2018-09-01 Michael Matz <matz@suse.de>
1819 PR tree-optimization/87074
1820 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1821 PHIs for outer-loop uses.
1823 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1825 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1826 * doc/invoke.texi (C Dialect Options): Ditto.
1828 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1830 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1832 2018-08-31 Richard Biener <rguenther@suse.de>
1834 PR tree-optimization/87168
1835 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1836 (rpo_elim::eliminate_avail): When OP was not visited it must
1839 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1841 * tree-vrp.c (copy_value_range): Convert param "from" from
1842 "value_range *" to "const value_range *".
1843 (range_is_null): Likewise for param "vr".
1844 (range_int_cst_p): Likewise.
1845 (range_int_cst_singleton_p): Likewise.
1846 (symbolic_range_p): Likewise.
1847 (value_ranges_intersect_p): Likewise for both params.
1848 (value_range_nonnegative_p): Likewise for param "vr".
1849 (value_range_constant_singleton): Likewise.
1850 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1851 (extract_range_into_wide_ints): Likewise for param "vr".
1852 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1854 (vrp_can_optimize_bit_op): Likewise.
1855 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1857 (extract_range_from_unary_expr): Likewise.
1858 (debug_value_range): Likewise for param "vr".
1859 (value_range::dump): Add "const" qualifier.
1860 (vrp_prop::check_array_ref): Convert local "vr" from
1861 "value_range *" to "const value_range *".
1862 (vrp_prop::check_mem_ref): Likewise.
1863 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1864 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1865 (vrp_intersect_ranges): Likewise.
1866 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1867 (vrp_prop::vrp_finalize): Likewise.
1868 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1869 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1870 (extract_range_from_unary_expr): Likewise.
1871 (value_range_constant_singleton): Likewise.
1872 (symbolic_range_p): Likewise.
1873 (copy_value_range): Likewise.
1874 (extract_range_from_binary_expr_1): Likewise.
1875 (range_int_cst_p): Likewise.
1876 (vrp_set_zero_nonzero_bits): Likewise.
1877 (range_int_cst_singleton_p): Likewise.
1879 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1881 * config/aarch64/arm_neon.h (vabsd_s64): New.
1882 (vnegd_s64): Likewise.
1884 2018-08-31 Martin Jambor <mjambor@suse.cz>
1886 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1888 2018-08-31 Martin Liska <mliska@suse.cz>
1890 * ipa-icf.c (sem_item::add_type): Use
1891 sem_item::m_type_hash_cache.
1892 * ipa-icf.h: Move the cache from sem_item_optimizer
1895 2018-08-31 Nathan Sidwell <nathan@acm.org>
1897 * doc/extend.texi (Backwards Compatibility): Remove implicit
1898 extern C leeway of () being (...).
1900 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1902 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1904 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1908 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1910 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1913 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1914 gen_int_mode. Formatting fixes.
1916 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1918 * target.def (custom_function_descriptors): Improve documentation.
1919 * doc/tm.texi.in (Trampolines): Expand discussion of function
1920 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1921 beginning of the section.
1922 * doc/tm.texi: Regenerated.
1924 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1926 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1927 base class in the initializer list. This is a workaround for
1928 building with older GCC.
1929 (class auto_bb_flag): Likewise.
1931 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1933 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1934 (altivec_vcmpequ<VI_char>_p): Remove star.
1935 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1937 (expand_strncmp_vec_sequence): New function.
1938 (emit_final_str_compare_vec): New function.
1939 (expand_strn_compare): Add support for vector strncmp.
1940 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1941 length specification to bytes.
1942 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1943 (vcmpnezb_p): New pattern.
1944 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1945 for option -mstring-compare-inline-limit.
1947 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1949 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1950 (PTA_SKYLAKE): Add PTA_AES.
1951 (PTA_GOLDMONT): Likewise.
1953 2018-08-29 Jan Hubicka <jh@suse.cz>
1956 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1957 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1959 2018-08-29 Jan Hubicka <jh@suse.cz>
1961 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1963 (hash_tree): Do not visit TYPE_STUB_DECL.
1964 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1965 stream TYPE_STUB_DECL.
1966 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1967 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1968 after free_lang_data.
1969 (type_in_anonymous_namespace_p): Likewise.
1971 2018-08-29 Jan Hubicka <jh@suse.cz>
1973 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1974 comment that it has to be even number.
1975 (class sreal): Change m_sig type to int32_t.
1976 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1977 int64_t for temporary calculations.
1978 (sreal_verify_basics): Drop one bit from minimum and maximum.
1980 2018-08-30 Richard Biener <rguenther@suse.de>
1982 PR tree-optimization/87147
1983 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1984 (visit_phi): When the degenerate result is from the backedge and
1985 we didn't visit its definition yet drop to VARYING.
1986 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1988 2018-08-29 Jan Hubicka <jh@suse.cz>
1990 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1992 (hash_tree): Likewise.
1994 2018-08-29 Jan Hubicka <jh@suse.cz>
1996 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1997 and TYPE_NEXT_REF_TO.
1999 2018-08-29 Jan Hubicka <jh@suse.cz>
2001 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2002 comment that it has to be even number.
2003 (class sreal): Change m_sig type to int32_t.
2004 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2005 int64_t for temporary calculations.
2006 (sreal_verify_basics): Drop one bit from minimum and maximum.
2008 2018-08-30 Tamar Christina <tamar.christina@arm.com>
2010 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
2012 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
2015 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
2016 if to_add is negative.
2018 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
2021 * builtins.c (c_strlen): Improve range checks.
2023 2018-08-29 Martin Sebor <msebor@redhat.com>
2024 Jeff Law <law@redhat.com>
2026 PR tree-optimization/86714
2027 PR tree-optimization/86711
2028 * builtins.c (c_strlen): Add arguments to call to string_constant.
2029 * expr.c (string_constant): Add argument. Detect missing nul
2030 terminator and outermost declaration it's missing in.
2031 * expr.h (string_constant): Add argument.
2032 * fold-const.c (read_from_constant_string): Add arguments to call to
2034 (c_getstr): Likewise.
2035 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2037 * tree-ssa-strlen.c (get_stridx): Likewise.
2039 2018-08-29 Jan Hubicka <jh@suse.cz>
2041 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
2042 Do not stream DECL_VINDEX.
2043 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
2044 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
2045 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
2047 2018-08-29 Richard Biener <rguenther@suse.de>
2049 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
2050 virtual operands that are not default defs to honor region
2052 (rpo_vn_valueize): Remove ineffective code here.
2054 2018-08-29 Richard Biener <rguenther@suse.de>
2056 PR tree-optimization/87132
2057 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
2058 when skipping defs reachable over backedges.
2060 2018-08-29 Richard Biener <rguenther@suse.de>
2062 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
2063 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
2064 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
2065 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
2066 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
2067 (vn_reference_lookup_3): Remove use of const_parms.
2068 (free_rpo_vn): Do not free const_parms.
2069 (do_rpo_vn): Do not call init_const_parms.
2070 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
2071 SSA_NAME_POINTS_TO_READONLY_MEMORY.
2072 (call_may_clobber_ref_p_1): Likewise.
2074 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
2077 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
2079 (-ftree-scev-cprop): Document.
2081 2018-08-29 Jan Hubicka <jh@suse.cz>
2083 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
2085 (sreal constructor): Update.
2086 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
2087 sreal:operator/): Update.
2089 2018-08-29 Martin Liska <mliska@suse.cz>
2091 * tree-switch-conversion.c (switch_conversion::expand):
2092 Strenghten assumption about gswitch statements.
2094 2018-08-29 Richard Biener <rguenther@suse.de>
2096 PR tree-optimization/87117
2097 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
2098 re-value-number released SSA VDEFs.
2100 2018-08-29 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/87126
2103 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
2105 2018-08-28 Jim Wilson <jimw@sifive.com>
2107 * config/riscv/pic.md: Rewrite.
2108 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
2110 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
2111 (SOFTF, default_load, softload, softstore): New.
2113 2018-08-28 Jeff Law <law@redhat.com>
2115 * fold-const.c (fold_binary_loc): Remove recently added assert.
2117 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
2119 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
2120 to OP parmeter of generated function.
2122 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
2124 PR tree-optimization/87009
2125 * match.pd: Add boolean optimizations.
2127 2018-08-28 Martin Sebor <msebor@redhat.com>
2130 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
2131 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
2132 (pass_walloca::gate): Use it.
2133 (alloca_call_type): Same.
2134 (pass_walloca::execute): Same.
2135 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
2137 2018-08-28 David Malcolm <dmalcolm@redhat.com>
2139 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
2140 GCC_VERSION for usage of "__gcc_dump_printf__" format from
2143 2018-08-28 Richard Biener <rguenther@suse.de>
2145 PR tree-optimization/87124
2146 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
2147 constants before looking up avail.
2149 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2152 * calls.c (maybe_warn_nonstring_arg): Punt early if
2153 warn_stringop_overflow is zero. Don't call get_range_strlen
2154 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2155 Swap comparison operands to have constants on rhs. Only use
2156 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2157 increment lenrng[0].
2159 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
2161 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
2162 use of tree_to_shwi. Remove duplicated test for the size being
2163 a whole number of bytes.
2165 2018-08-28 Richard Biener <rguenther@suse.de>
2167 PR tree-optimization/87117
2168 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
2169 Handle removed stmt without LHS (GIMPLE_NOP).
2171 2018-08-28 Richard Biener <rguenther@suse.de>
2173 PR tree-optimization/87117
2174 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
2175 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
2177 2018-08-28 Richard Biener <rguenther@suse.de>
2179 PR tree-optimization/87117
2180 * tree-ssa-pre.c (compute_avail): Do not make expressions
2181 with predicated values available.
2182 (get_expr_value_id): Assert we do not run into predicated value
2185 2018-08-28 Richard Biener <rguenther@suse.de>
2187 PR tree-optimization/87117
2188 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
2189 get virtual operands.
2190 (get_expr_operands): Handle STRING_CST like other decls.
2192 2018-08-28 Martin Liska <mliska@suse.cz>
2194 * tree.h: Update documentation of fndecl_built_in_p
2198 2018-08-27 Jeff Law <law@redhat.com>
2199 PR tree-optimization/87110
2200 * tree-ssa-dse.c (compute_trims): Handle non-constant
2203 2018-08-27 Martin Sebor <msebor@redhat.com>
2205 PR tree-optimization/86914
2206 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2208 2018-08-27 Martin Sebor <msebor@redhat.com>
2210 PR tree-optimization/87112
2211 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
2212 the type of the bound argument.
2214 2018-08-27 Jeff Law <law@redhat.com>
2216 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
2217 type does not have a TYPE_SIZE_UNIT.
2219 2018-08-27 Steve Ellcey <sellcey@cavium.com>
2221 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
2222 with include of backend.h.
2224 2018-08-27 Richard Biener <rguenther@suse.de>
2226 PR tree-optimization/86927
2227 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2228 use const cond reduction code.
2230 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
2232 PR tree-optimization/85758
2233 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
2235 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2238 * diagnostic-show-locus.c (class layout_range): Update for
2239 conversion of show_caret_p to a tri-state.
2240 (layout_range::layout_range): Likewise.
2241 (make_range): Likewise.
2242 (layout::maybe_add_location_range): Likewise.
2243 (layout::should_print_annotation_line_p): Don't show annotation
2244 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
2245 (layout::get_state_at_point): Update for conversion of
2246 show_caret_p to a tri-state. Bail out early for
2247 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
2248 underlining or source colorization.
2249 (gcc_rich_location::add_location_if_nearby): Update for conversion
2250 of show_caret_p to a tri-state.
2251 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
2252 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
2254 (selftest::test_one_liner_labels): Likewise.
2255 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
2256 conversion of show_caret_p to a tri-state.
2257 * pretty-print.c (text_info::set_location): Likewise.
2258 * pretty-print.h (text_info::set_location): Likewise.
2259 * substring-locations.c (format_warning_n_va): Likewise.
2260 * tree-diagnostic.c (default_tree_printer): Likewise.
2261 * tree-pretty-print.c (newline_and_indent): Likewise.
2263 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2266 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
2267 line above for line-insertion fix-it hints.
2268 (selftest::test_fixit_insert_containing_newline): Update the
2269 expected results, and add a test with line-numbering enabled.
2271 2018-08-27 Martin Liska <mliska@suse.cz>
2274 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
2275 params with DECL_HAS_VALUE_EXPR_P.
2277 2018-08-27 Martin Liska <mliska@suse.cz>
2279 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
2280 selected expansion strategy.
2282 2018-08-27 Martin Liska <mliska@suse.cz>
2284 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
2285 * builtins.c (is_builtin_fn): Likewise.
2286 * attribs.c (diag_attr_exclusions): Use new function
2287 fndecl_built_in_p and remove check for FUNCTION_DECL if
2289 (builtin_mathfn_code): Likewise.
2290 (fold_builtin_expect): Likewise.
2291 (fold_call_expr): Likewise.
2292 (fold_builtin_call_array): Likewise.
2293 (fold_call_stmt): Likewise.
2294 (set_builtin_user_assembler_name): Likewise.
2295 (is_simple_builtin): Likewise.
2296 * calls.c (gimple_alloca_call_p): Likewise.
2297 (maybe_warn_nonstring_arg): Likewise.
2298 * cfgexpand.c (expand_call_stmt): Likewise.
2299 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
2300 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
2301 (cgraph_node::verify_node): Likewise.
2302 * cgraphclones.c (build_function_decl_skip_args): Likewise.
2303 (cgraph_node::create_clone): Likewise.
2304 * config/arm/arm.c (arm_insert_attributes): Likewise.
2305 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
2306 * dse.c (scan_insn): Likewise.
2307 * expr.c (expand_expr_real_1): Likewise.
2308 * fold-const.c (operand_equal_p): Likewise.
2309 (fold_binary_loc): Likewise.
2310 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
2311 * gimple-low.c (lower_stmt): Likewise.
2312 * gimple-pretty-print.c (dump_gimple_call): Likewise.
2313 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
2315 * gimple.c (gimple_build_call_from_tree): Likewise.
2316 (gimple_call_builtin_p): Likewise.
2317 (gimple_call_combined_fn): Likewise.
2318 * gimplify.c (gimplify_call_expr): Likewise.
2319 (gimple_boolify): Likewise.
2320 (gimplify_modify_expr): Likewise.
2321 (gimplify_addr_expr): Likewise.
2322 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
2323 * ipa-cp.c (determine_versionability): Likewise.
2324 * ipa-fnsummary.c (compute_fn_summary): Likewise.
2325 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
2326 * ipa-split.c (visit_bb): Likewise.
2327 (split_function): Likewise.
2328 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2329 * lto-cgraph.c (input_node): Likewise.
2330 * lto-streamer-out.c (write_symbol): Likewise.
2331 * omp-low.c (setjmp_or_longjmp_p): Likewise.
2332 (lower_omp_1): Likewise.
2333 * predict.c (strip_predict_hints): Likewise.
2334 * print-tree.c (print_node): Likewise.
2335 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
2336 * trans-mem.c (is_tm_irrevocable): Likewise.
2337 (is_tm_load): Likewise.
2338 (is_tm_simple_load): Likewise.
2339 (is_tm_store): Likewise.
2340 (is_tm_simple_store): Likewise.
2341 (is_tm_abort): Likewise.
2342 (tm_region_init_1): Likewise.
2343 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
2344 * tree-cfg.c (verify_gimple_call): Likewise.
2345 (move_stmt_r): Likewise.
2346 (stmt_can_terminate_bb_p): Likewise.
2347 * tree-eh.c (lower_eh_constructs_2): Likewise.
2348 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2349 * tree-inline.c (remap_gimple_stmt): Likewise.
2350 (copy_bb): Likewise.
2351 (estimate_num_insns): Likewise.
2352 (fold_marked_statements): Likewise.
2353 * tree-sra.c (scan_function): Likewise.
2354 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
2355 (optimize_stack_restore): Likewise.
2356 (pass_fold_builtins::execute): Likewise.
2357 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
2358 (mark_all_reaching_defs_necessary_1): Likewise.
2359 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
2360 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2361 (pass_forwprop::execute): Likewise.
2362 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2363 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2364 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
2365 * tree-ssa-strlen.c (get_string_length): Likewise.
2366 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
2367 (find_func_aliases_for_call): Likewise.
2368 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
2369 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
2370 * tree-tailcall.c (find_tail_calls): Likewise.
2371 * tree.c (need_assembler_name_p): Likewise.
2372 (free_lang_data_in_decl): Likewise.
2373 (get_call_combined_fn): Likewise.
2374 * ubsan.c (is_ubsan_builtin_p): Likewise.
2375 * varasm.c (incorporeal_function_p): Likewise.
2376 * tree.h (DECL_BUILT_IN): Remove and replace with
2378 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
2379 (fndecl_built_in_p): New.
2381 2018-08-27 Martin Liska <mliska@suse.cz>
2383 PR tree-optimization/86847
2384 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
2385 Dump also subtree probability.
2386 (switch_decision_tree::do_jump_if_equal): New function.
2387 (switch_decision_tree::emit_case_nodes): Handle special
2388 situations in balanced tree that can be emitted much simpler.
2389 Fix calculation of probabilities that happen in tree expansion.
2390 * tree-switch-conversion.h (struct cluster): Add
2392 (struct simple_cluster): Likewise.
2393 (struct case_tree_node): Add new function has_child.
2394 (do_jump_if_equal): New.
2396 2018-08-27 Martin Liska <mliska@suse.cz>
2398 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
2399 Add new argument to bit_test_cluster constructor.
2400 (bit_test_cluster::emit): Set bits really number of values
2402 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
2403 probability argument.
2404 * tree-switch-conversion.h (struct bit_test_cluster):
2405 Add m_handles_entire_switch.
2407 2018-08-27 Martin Liska <mliska@suse.cz>
2409 PR tree-optimization/86702
2410 * tree-switch-conversion.c (jump_table_cluster::emit):
2411 Make probabilities even for values in jump table
2412 according to number of cases handled.
2413 (switch_decision_tree::compute_cases_per_edge): Pass
2414 argument to reset_out_edges_aux function.
2415 (switch_decision_tree::analyze_switch_statement): Likewise.
2416 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
2419 2018-08-27 Martin Liska <mliska@suse.cz>
2421 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
2422 cfun argument explicitly.
2423 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
2424 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
2425 function gimple_switch_default_bb.
2426 (convert_switch_statements):
2428 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
2429 * stmt.c (label_to_block_fn): Use label_to_block and pass
2430 cfun argument explicitly and use gimple_switch_label_bb.
2431 (expand_case): Likewise.
2432 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
2433 cfun argument explicitly. Likewise.
2434 (make_edges_bb): Likewise.
2435 (make_cond_expr_edges): Likewise.
2436 (get_cases_for_edge): Likewise.
2437 (make_gimple_switch_edges): Likewise.
2438 (label_to_block_fn): Likewise.
2439 (label_to_block): Likewise.
2440 (make_goto_expr_edges): Likewise.
2441 (make_gimple_asm_edges): Likewise.
2442 (main_block_label): Likewise.
2443 (group_case_labels_stmt): Likewise.
2444 (find_taken_edge_computed_goto): Likewise.
2445 (find_taken_edge_switch_expr): Likewise.
2446 (gimple_verify_flow_info): Likewise.
2447 (gimple_redirect_edge_and_branch): Likewise.
2448 (gimple_switch_label_bb): New function.
2449 (gimple_switch_default_bb): Likewise.
2450 (gimple_switch_edge): Likewise.
2451 (gimple_switch_default_edge): Likewise.
2452 * tree-cfg.h (label_to_block_fn): Remove and replace ...
2453 (label_to_block): ... with this.
2454 (gimple_switch_label_bb): New.
2455 (gimple_switch_default_bb): Likewise.
2456 (gimple_switch_edge): Likewise.
2457 (gimple_switch_default_edge): Likewise.
2458 * tree-cfgcleanup.c (convert_single_case_switch): Use
2459 new gimple functions and pass new argument to label_to_block.
2460 (cleanup_control_flow_bb):
2461 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2462 cfun argument explicitly.
2463 (make_eh_edges): Likewise.
2464 (redirect_eh_dispatch_edge): Likewise.
2465 (lower_resx): Likewise.
2466 (lower_eh_dispatch): Likewise.
2467 (maybe_remove_unreachable_handlers): Likewise.
2468 (unsplit_eh): Likewise.
2469 (cleanup_empty_eh): Likewise.
2470 (verify_eh_edges): Likewise.
2471 (verify_eh_dispatch_edge): Likewise.
2472 * tree-ssa-dom.c (record_edge_info): Likewise.
2473 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2474 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2475 (thread_through_normal_block): Likewise.
2476 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2477 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2478 * tree-switch-conversion.c (switch_conversion::collect): Use new
2480 (switch_conversion::check_final_bb): Likewise.
2481 (switch_conversion::gather_default_values): Pass new argument
2483 (switch_conversion::build_constructors): Likewise.
2484 (switch_decision_tree::compute_cases_per_edge): Use new
2485 gimple_switch_edge function.
2486 (switch_decision_tree::analyze_switch_statement): Pass new argument
2488 (switch_decision_tree::try_switch_expansion): Use
2489 gimple_switch_default_edge.
2490 * tree-vrp.c (find_switch_asserts): Pass new argument
2492 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2493 (vr_values::simplify_switch_using_ranges): Likewise.
2495 2018-08-27 Richard Biener <rguenther@suse.de>
2497 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2498 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2500 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2501 (struct vn_nary_op_s): Add unwind_to member. Add
2502 predicated_values flag and put result into a union together
2503 with a linked list of vn_pval.
2504 (struct vn_ssa_aux): Add name member to make maintaining
2505 a map of SSA name to vn_ssa_aux possible. Remove no longer
2506 needed info, dfsnum, low, visited, on_sccstack, use_processed
2507 and range_info_anti_range_p members.
2508 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2509 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2511 (vn_valueize): New global.
2512 (vn_context_bb): Likewise.
2513 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2514 VN_INFO_PTR_INFO): Remove.
2515 * tree-ssa-sccvn.c: ... (rewrite)
2516 (pass_fre::execute): For -O2+ initialize loops and run
2517 RPO VN in optimistic mode (iterating). For -O1 and -Og
2518 run RPO VN in non-optimistic mode.
2519 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2520 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2521 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2522 (rpo-vn-max-loop-depth): Document.
2523 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2524 when valuezing the VUSE signals we walked out of the region.
2525 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2526 (phi_translate): Set VN context block to use for availability
2528 (compute_avail): Likewise.
2529 (pre_valueize): New function.
2530 (pass_pre::execute): Adjust to the RPO VN API.
2532 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2533 (propagate_constants_for_unrolling): Remove.
2534 (tree_unroll_loops_completely): Perform value-numbering
2535 on the unrolled bodies loop parent.
2537 2018-08-27 Richard Biener <rguenther@suse.de>
2539 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2540 for partial antic compute.
2542 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2544 PR rtl-optimization/87065
2545 * combine.c (simplify_if_then_else): Formatting fix.
2546 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2548 (known_cond): Don't return const_true_rtx for vector modes. Use
2549 CONST0_RTX instead of const0_rtx. Formatting fixes.
2551 2018-08-27 Martin Liska <mliska@suse.cz>
2553 PR gcov-profile/87069
2554 * gcov.c (process_file): Record files already processed
2555 and warn about a file being processed multiple times.
2557 2018-08-27 Martin Liska <mliska@suse.cz>
2560 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2561 Set default values for x_aarch64_*_string strings.
2562 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2563 prefix. For -mabi do not print '=ABI' in help and use
2564 <option_value> format for -msve-vector-bits and -moverride
2567 2018-08-26 Jeff Law <law@redhat.com>
2569 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2572 2018-08-26 Marek Polacek <polacek@redhat.com>
2574 PR c++/87029, Implement -Wredundant-move.
2575 * doc/invoke.texi: Document -Wredundant-move.
2577 2018-08-25 Martin Sebor <msebor@redhat.com>
2579 PR tree-optimization/87059
2580 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2581 to the same type as the other.
2582 * fold-const.c (fold_binary_loc): Assert expectation.
2584 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2586 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2587 extraneous parentheses, dead code section and formatting.
2589 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2592 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2593 wide enough for jumps in the line-numbering to be visible.
2594 (layout::print_gap_in_line_numbering): New member function.
2595 (layout::calculate_line_spans): When using line numbering, merge
2596 line spans that are only 1 line apart.
2597 (diagnostic_show_locus): When printing line numbers, show gaps in
2598 line numbering directly, rather than printing headers.
2599 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2600 line-numbering with multiple line spans.
2601 (selftest::test_fixit_insert_containing_newline_2): Add test of
2602 line-numbering, in which the spans are close enough to be merged.
2604 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2606 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2607 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2608 * tree-vrp.c (range_is_nonnull): Remove.
2609 (range_includes_zero_p): Accept value_range instead of min/max.
2610 (extract_range_from_binary_expr_1): Do not early bail on
2612 Use range_includes_zero_p instead of range_is_nonnull.
2613 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2614 of range_is_nonnull.
2615 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2616 special case VR_ANTI_RANGE.
2617 (vrp_finalize): Same.
2618 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2620 (range_is_nonnull): Remove.
2621 * vr-values.c (vrp_stmt_computes_nonzero): Use
2622 range_includes_zero_p instead of range_is_nonnull.
2623 (extract_range_basic): Pass value_range to range_includes_zero_p
2624 instead of range_is_nonnull.
2626 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2628 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2629 * emit-rtl.h (rtl_data): Remove return_bnd.
2630 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2631 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2632 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2633 (POINTER_BOUNDS_MODE): Remove definition.
2634 (make_pointer_bounds_mode): Remove.
2635 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2636 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2637 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2638 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2639 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2640 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2641 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2643 * config/i386/i386-modes.def (BND32, BND64): Remove.
2644 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2645 (dbx64_register_map): Ditto.
2646 (svr4_dbx_register_map): Ditto.
2647 (indirect_thunk_bnd_needed): Remove.
2648 (indirect_thunks_bnd_used): Ditto.
2649 (indirect_return_bnd_needed): Ditto.
2650 (indirect_return_via_cx_bnd): Ditto.
2651 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2652 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2653 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2654 (output_indirect_thunk_function): Remove handling of
2655 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2656 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2657 (ix86_save_reg): Remove handling of crtl->return_bnd.
2658 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2659 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2660 and UNSPEC_BNDLX_ADDR.
2661 (ix86_output_indirect_branch_via_reg): Remove handling of
2662 indirect_thunk_prefix_bnd.
2663 (ix86_output_indirect_branch_via_push): Ditto.
2664 (ix86_output_function_return): Ditto.
2665 (ix86_output_indirect_function_return): Ditto.
2666 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2667 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2668 (CALL_USED_REGISTERS): Ditto.
2669 (REG_ALLOC_ORDER): Update for removal of bound registers.
2670 (HI_REGISTER_NAMES): Ditto.
2671 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2672 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2673 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2674 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2675 (FIRST_PSEUDO_REG): Update.
2676 (BND): Remove mode iterator.
2677 * config/i386/predicates.md (bnd_mem_operator): Remove.
2679 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2681 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2684 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2686 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2687 the case in which the permute needs only a single element and
2688 repeats for every vector of the result. Extend that case to
2689 handle variable-length vectors.
2690 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2692 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
2695 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2696 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2698 2018-08-24 Richard Biener <rguenther@suse.de>
2700 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2701 bb_flags_allocated members.
2702 (auto_flag): New RAII class for allocating flags.
2703 (auto_edge_flag): New RAII class for allocating edge flags.
2704 (auto_bb_flag): New RAII class for allocating bb flags.
2705 * cfgloop.c (verify_loop_structure): Allocate temporary edge
2707 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2708 in favor of temporarily allocated BB flag.
2709 * hsa-brig.c: Re-order includes.
2710 * hsa-dump.c: Likewise.
2711 * hsa-regalloc.c: Likewise.
2712 * print-rtl.c: Likewise.
2713 * profile-count.c: Likewise.
2715 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2718 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2721 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2724 * wide-int-range.cc (wide_int_range_div): Do not ignore result
2725 from wide_int_range_multiplicative_op.
2727 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2729 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2732 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
2734 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2737 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
2739 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2740 full GENERIC expression used for replacement.
2742 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
2744 * tree-vrp.c (abs_extent_range): Remove.
2745 (extract_range_into_wide_ints): Pass wide ints by reference.
2746 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2747 Pass wide ints by reference in all calls to
2748 extract_range_into_wide_ints.
2749 * wide-int-range.cc (wide_int_range_div): New.
2750 * wide-int-range.h (wide_int_range_div): New.
2751 (wide_int_range_includes_zero_p): New.
2752 (wide_int_range_zero_p): New.
2754 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
2756 * config/aarch64/aarch64.md (arches): New enum.
2757 (arch): New enum attr.
2758 (arch_enabled): New attr.
2759 (enabled): Now uses arch_enabled only.
2760 (simd, sve, fp16): Removed attribute.
2761 (fp): Attr now defined in terms of 'arch'.
2762 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2763 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2764 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2765 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2766 attributes into 'arch'.
2767 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2768 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2769 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2770 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2771 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2772 'simd' attribute into 'arch'.
2773 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2774 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2775 Convert use of 'fp' attribute to 'arch'.
2776 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2777 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2779 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2780 (different modes) Merge 'fp' and 'simd' into 'arch'.
2781 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2784 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
2786 PR rtl-optimization/87026
2787 * expmed.c (canonicalize_comparison): If we can no longer create
2788 pseudoregisters, don't.
2790 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
2793 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2795 * config/arm/arm.c (speculation_barrier_libfunc): New static
2797 (arm_init_libfuncs): Initialize it.
2798 (arm_emit_speculation_barrier): New function.
2799 * config/arm/arm.md (speculation_barrier): Call
2800 arm_emit_speculation_barrier for architectures that do not have
2802 (speculation_barrier_insn): Only match on Armv7 or later.
2804 2018-08-23 Richard Biener <rguenther@suse.de>
2807 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2810 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2812 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2813 of single-vector TBLs.
2814 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2817 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2820 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2821 aarch64_evpc_tbl guard.
2823 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2825 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2828 2018-08-22 Martin Sebor <msebor@redhat.com>
2831 * tree-pretty-print.c (pretty_print_string): Add argument.
2832 (dump_generic_node): Call to pretty_print_string with string size.
2834 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2836 PR rtl-optimization/86771
2837 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2838 of two SETs into those two SETs, one to be placed at i2, if that SETs
2839 destination is modified between i2 and i3.
2841 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2843 PR tree-optimization/86725
2844 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2846 (vect_analyze_scalar_cycles_1): Check it.
2848 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2850 PR tree-optimization/86725
2851 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2852 an outer loop phi as a double reduction, make sure that the
2853 single user of the phi result is an inner loop phi.
2855 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2857 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2858 grouped stores with gaps to a strided group.
2860 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2862 * tree-vect-stmts.c (get_group_load_store_type)
2863 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2864 first statement in a group.
2866 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2868 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2869 the sequence used in gcc/gcc.c.
2871 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2874 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2877 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2879 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2880 Darwin10-specific unwinder-shim.
2881 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2882 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2883 New to cater for Darwin10 Rosetta.
2885 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2887 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2888 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2890 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2895 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2896 (dwarf2out_switch_text_section): Generate a local label for the
2897 second function sub-section and apply it as the second FDE start
2899 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2900 second sub-section start.
2902 2018-08-22 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/86988
2905 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2907 2018-08-22 Richard Biener <rguenther@suse.de>
2909 PR tree-optimization/86945
2910 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2912 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2914 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2915 a comment about how uses of r2 for .sdata2 come about.
2917 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2919 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2921 2018-08-21 Marek Polacek <polacek@redhat.com>
2923 PR c++/86981, Implement -Wpessimizing-move.
2924 * doc/invoke.texi: Document -Wpessimizing-move.
2926 2018-08-21 Jan Hubicka <jh@suse.cz>
2928 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2929 * tree.h (is_redundant_typedef): Remove.
2930 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2932 2018-08-21 Jan Hubicka <jh@suse.cz>
2934 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2937 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2939 * expmed.c (extract_low_bits): Reject invalid subregs early.
2941 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2944 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2947 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2949 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2950 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2951 * config/vxworks.c: Likewise.
2953 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2955 * config/vxworks.c: Set targetm.have_ctors_dtors
2956 if HAVE_INITFINI_ARRAY_SUPPORT.
2957 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2958 if HAVE_INITFINI_ARRAY_SUPPORT.
2960 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2962 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2963 default search path for VxWorks < 7.
2965 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2967 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2968 (get_format_string): Refer to c_getstr.
2970 2018-08-21 Tom de Vries <tdevries@suse.de>
2972 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2973 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2974 (debuginfo_early_stop): Declare.
2975 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2976 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2977 (debuginfo_early_stop): New function.
2978 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2979 and debuginfo_early_stop.
2980 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2981 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2982 (general_init): Call debuginfo_early_init.
2983 (finalize): Call debuginfo_fini.
2984 (do_compile): Call debuginfo_init.
2985 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2987 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2989 2018-08-21 Tom de Vries <tdevries@suse.de>
2991 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2992 flag_dump_noaddr and flag_dump_unnumbered.
2994 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2996 * wide-int-range.cc (wide_int_range_abs): New.
2997 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2998 * wide-int-range.h (wide_int_range_abs): New.
2999 (wide_int_range_min_max): New.
3000 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
3001 case to call wide_int_range_abs.
3002 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
3003 (extract_range_from_abs_expr): Delete.
3005 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
3008 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3009 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3012 2018-08-20 Nathan Sidwell <nathan@acm.org>
3013 Jeff Law <law@redhat.com>
3015 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
3016 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
3018 2018-08-20 David Malcolm <dmalcolm@redhat.com>
3021 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
3022 (decl_attributes): Likewise.
3023 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
3025 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3026 * diagnostic-core.h (class auto_diagnostic_group): New class.
3027 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
3028 (diagnostic_report_diagnostic): Handle the first diagnostics within
3030 (emit_diagnostic): Add auto_diagnostic_group instance.
3032 (inform_n): Likewise.
3033 (warning): Likewise.
3034 (warning_at): Likewise.
3035 (warning_n): Likewise.
3036 (pedwarn): Likewise.
3037 (permerror): Likewise.
3039 (error_n): Likewise.
3040 (error_at): Likewise.
3042 (fatal_error): Likewise.
3043 (internal_error): Likewise.
3044 (internal_error_no_backtrace): Likewise.
3045 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
3046 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
3047 * diagnostic.h (struct diagnostic_context): Add fields
3048 "diagnostic_group_nesting_depth",
3049 "diagnostic_group_emission_count", "begin_group_cb",
3051 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
3052 Add auto_diagnostic_group instance(s).
3053 (find_explicit_erroneous_behavior): Likewise.
3054 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
3055 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
3056 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
3057 (gimplify_va_arg_expr): Likewise.
3058 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
3059 (HSA_SORRY_AT): Likewise.
3060 * ipa-devirt.c (compare_virtual_tables): Likewise.
3061 (warn_odr): Likewise.
3062 * multiple_target.c (expand_target_clones): Likewise.
3063 * opts-common.c (cmdline_handle_error): Likewise.
3064 * reginfo.c (globalize_reg): Likewise.
3065 * substring-locations.c (format_warning_n_va): Likewise.
3066 * tree-inline.c (expand_call_inline): Likewise.
3067 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
3068 * tree-ssa-loop-niter.c
3069 (do_warn_aggressive_loop_optimizations): Likewise.
3070 * tree-ssa-uninit.c (warn_uninit): Likewise.
3071 * tree.c (warn_deprecated_use): Likewise.
3073 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
3076 * config/i386/i386.md (eh_return): Always update EH return
3077 address in word_mode.
3079 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
3081 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
3082 TARGET_SPLIT_COMPLEX_ARG is defined.
3084 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3086 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
3088 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3091 * expr.c (expand_assignment): Assert that bitpos is positive.
3092 (store_field): Likewise
3093 (expand_expr_real_1): Make sure that bitpos is positive.
3094 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
3097 2018-08-20 Nathan Sidwell <nathan@acm.org>
3099 * Makefile.in (CPP_ID_DATA_H): Delete.
3100 (CPP_INTERNAL_H): Don't add it.
3101 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
3102 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
3104 2018-08-20 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/78655
3107 * tree-vrp.c (extract_range_from_binary_expr_1): Make
3108 pointer + offset nonnull if either operand is nonnull work.
3110 2018-08-20 Tom de Vries <tdevries@suse.de>
3112 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
3113 unless the referenced die describes the added property using
3114 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
3115 Otherwise, add a DW_AT_location to the referenced die.
3117 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
3120 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
3121 register_operand when calling ix86_set_reg_reg_cost.
3122 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
3123 Set *total to 0 for operands that satisfy x86_64_immediate_operand
3124 predicate and to 1 otherwise.
3126 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
3128 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
3129 emit a diagnostic that it is not supported and reset the option.
3130 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
3131 supported and consume the option. (ASM_FINAL_SPEC): New.
3133 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3135 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
3138 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
3140 C-SKY port: Documentation
3142 * doc/extend.texi (C-SKY Function Attributes): New section.
3143 * doc/invoke.texi (Option Summary): Add C-SKY options.
3144 (C-SKY Options): New section.
3145 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
3147 2018-08-17 Jojo <jijie_rong@c-sky.com>
3148 Huibin Wang <huibin_wang@c-sky.com>
3149 Sandra Loosemore <sandra@codesourcery.com>
3150 Chung-Lin Tang <cltang@codesourcery.com>
3152 C-SKY port: Backend implementation
3154 * config/csky/*: New.
3155 * common/config/csky/*: New.
3157 2018-08-17 Jojo <jijie_rong@c-sky.com>
3158 Huibin Wang <huibin_wang@c-sky.com>
3159 Sandra Loosemore <sandra@codesourcery.com>
3160 Chung-Lin Tang <cltang@codesourcery.com>
3161 Andrew Jenner <andrew@codesourcery.com>
3163 C-SKY port: Configury
3165 * config.gcc (csky-*-*): New.
3166 * configure.ac: Add csky to targets for dwarf2 debug_line support.
3167 * configure: Regenerated.
3169 2018-08-17 David Malcolm <dmalcolm@redhat.com>
3171 * dump-context.h: Include "dumpfile.h".
3172 (dump_context::dump_printf_va): Convert final param from va_list
3173 to va_list *. Convert from ATTRIBUTE_PRINTF to
3174 ATTRIBUTE_GCC_DUMP_PRINTF.
3175 (dump_context::dump_printf_loc_va): Likewise.
3176 * dumpfile.c: Include "stringpool.h".
3177 (make_item_for_dump_printf_va): Delete.
3178 (make_item_for_dump_printf): Delete.
3179 (class dump_pretty_printer): New class.
3180 (dump_pretty_printer::dump_pretty_printer): New ctor.
3181 (dump_pretty_printer::emit_items): New member function.
3182 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
3184 (dump_pretty_printer::emit_item): New member function.
3185 (dump_pretty_printer::stash_item): New member function.
3186 (dump_pretty_printer::format_decoder_cb): New member function.
3187 (dump_pretty_printer::decode_format): New member function.
3188 (dump_context::dump_printf_va): Reimplement in terms of
3189 dump_pretty_printer.
3190 (dump_context::dump_printf_loc_va): Convert final param from va_list
3192 (dump_context::begin_scope): Reimplement call to
3193 make_item_for_dump_printf.
3194 (dump_printf): Update for change to dump_printf_va.
3195 (dump_printf_loc): Likewise.
3196 (selftest::test_capture_of_dump_calls): Convert "stmt" from
3197 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
3198 with %T, %E, and %G.
3199 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
3200 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
3201 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
3202 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
3203 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
3204 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
3205 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
3206 within a dump_printf_loc call to "%wu".
3207 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
3208 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
3209 missing space after "=".
3210 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
3211 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
3212 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
3213 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
3214 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
3215 duplicate "vectorized" from message.
3217 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
3219 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
3220 polyNxK_t element's TYPE_STRING_FLAG.
3222 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3224 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
3225 (they were unnamed before). Fix comments.
3227 2018-08-17 Nathan Sidwell <nathan@acm.org>
3229 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
3231 2018-08-17 Richard Biener <rguenther@suse.de>
3233 PR tree-optimization/86841
3234 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
3236 2018-08-17 Martin Liska <mliska@suse.cz>
3238 * common.opt: Remove Warn, Init and Report for options with
3239 Ignore/Deprecated flag. Warning is done automatically for
3241 * config/i386/i386.opt: Likewise.
3242 * config/ia64/ia64.opt: Likewise.
3243 * config/rs6000/rs6000.opt: Likewise.
3244 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
3245 Remove usage of flag_check_pointer_bounds.
3246 * lto-wrapper.c (merge_and_complain): Do not handle
3247 OPT_fcheck_pointer_bounds.
3248 (append_compiler_options): Likewise.
3249 * opt-functions.awk: Do not handle Deprecated.
3250 * optc-gen.awk: Check that Var, Report and Init are not
3251 used for an option with Ignore/Deprecated flag.
3252 * opts-common.c (decode_cmdline_option): Do not report
3254 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
3256 * opts.h (struct cl_option): Remove cl_deprecated flag.
3257 (CL_ERR_DEPRECATED): Remove error enum value.
3259 2018-08-17 Richard Biener <rguenther@suse.de>
3262 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3263 across a va-arg-pack using call adjust its return value accordingly.
3265 2018-08-16 Martin Sebor <msebor@redhat.com>
3267 PR tree-optimization/86853
3268 * gimple-ssa-sprintf.c (struct format_result): Rename member.
3269 (struct fmtresult): Add member and initialize it in ctors.
3270 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
3271 (format_string): Handle %S the same as %ls. Set MAYFAIL.
3272 (format_directive): Set POSUNDER4K when MAYFAIL is set.
3273 (parse_directive): Handle %C same as %c.
3274 (sprintf_dom_walker::compute_format_length): Adjust.
3275 (is_call_safe): Adjust.
3277 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3279 * builtins.c (c_strlen): Add new parameter eltsize. Use it
3280 for determining how to count the elements.
3281 * builtins.h (c_strlen): Adjust prototype.
3282 * expr.c (string_constant): Add new parameter mem_size.
3283 Set *mem_size appropriately.
3284 * expr.h (string_constant): Adjust protoype.
3285 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
3286 * gimple-fold.h (get_range_strlen): Adjust prototype.
3287 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
3288 (format_string): Call get_string_length with eltsize.
3290 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3292 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
3293 to emit the span, rather than setting it as the prefix.
3295 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3297 * diagnostic-show-locus.c (layout::start_annotation_line): Add
3298 "margin_char" parameter, defaulting to space. Use it in place
3299 of pp_space for the initial part of the margin.
3300 (layout::print_leading_fixits): Use '+' when filling the margin
3301 of line-insertion fix-it hints.
3303 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3305 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
3308 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3310 * config/rs6000/altivec.md: Don't set length attribute to the default
3312 * config/rs6000/darwin.md: Ditto.
3313 * config/rs6000/dfp.md: Ditto.
3314 * config/rs6000/htm.md: Ditto.
3315 * config/rs6000/rs6000.md: Ditto.
3316 * config/rs6000/sync.md: Ditto.
3317 * config/rs6000/vsx.md: Ditto.
3319 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3321 * config/rs6000/altivec.md: Don't set length attribute to the default
3322 value, for branch instructions.
3323 * config/rs6000/darwin.md: Ditto.
3324 * config/rs6000/rs6000.md: Ditto.
3326 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3328 * config/rs6000/rs6000.md (length): Always define as const_int 4.
3329 (unnamed conditional branch define_insn): Set length to 4 or 8
3330 depending on offset.
3331 (<bd>_<mode>): Similar, for alternative 0.
3332 (<bd>tf_<mode>): Ditto.
3334 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3336 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
3338 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
3340 * doc/rtl.texi: Replace old RTX class names with new names.
3343 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
3345 * expmed.h (canonicalize_comparison): New declaration.
3346 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
3347 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
3348 * optabs.c (prepare_cmp_insn): Likewise.
3349 * rtl.h (unsigned_condition_p): New function which checks if a
3350 comparison operator is unsigned.
3352 2018-08-16 Nathan Sidwell <nathan@acm.org>
3354 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
3355 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
3357 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3360 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
3361 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
3362 (mov<mov>): ..this and enable unconditionally.
3364 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3366 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
3368 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
3370 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
3371 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
3372 "Common" with "Target".
3374 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
3376 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
3377 * doc/invoke.texi (mmitigate-rop): Remove.
3378 * config/i386/i386.c: Do not include "regrename.h".
3379 (ix86_rop_should_change_byte_p, reg_encoded_number)
3380 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
3382 (ix86_reorg): Remove call to ix86_mitigate_rop.
3383 * config/i386/i386.md (attr "modrm_class"): Remove.
3384 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
3385 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
3386 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
3388 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
3390 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
3391 allow folding of mergeh() and mergel() for the float and double types.
3392 (fold_mergehl_helper): Rework to handle building a permute tree
3395 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
3397 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
3400 2018-08-15 David Malcolm <dmalcolm@redhat.com>
3402 * common.opt (fdiagnostics-show-labels): New option.
3403 * diagnostic-show-locus.c (class layout_range): Add field
3405 (class layout): Add field "m_show_labels_p".
3406 (layout_range::layout_range): Add param "label" and use it to
3408 (make_range): Pass in NULL for new "label" param of layout_range's
3410 (layout::layout): Initialize m_show_labels_p.
3411 (layout::maybe_add_location_range): Pass in loc_range->m_label
3412 when constructing layout_range instances.
3413 (struct line_label): New struct.
3414 (layout::print_any_labels): New member function.
3415 (layout::print_line): Call it if label-printing is enabled.
3416 (selftest::test_one_liner_labels): New test.
3417 (selftest::test_diagnostic_show_locus_one_liner): Call it.
3418 * diagnostic.c (diagnostic_initialize): Initialize
3419 context->show_labels_p.
3420 * diagnostic.h (struct diagnostic_context): Add field
3422 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3423 -fno-diagnostics-show-labels.
3424 * dwarf2out.c (gen_producer_string): Add
3425 OPT_fdiagnostics_show_labels to the ignored options.
3426 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
3428 (gcc_rich_location::maybe_add_expr): Likewise.
3429 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
3430 label" param, defaulting to NULL.
3431 (gcc_rich_location::add_expr): Add "label" param.
3432 (gcc_rich_location::maybe_add_expr): Likewise.
3433 (class text_range_label): New class.
3434 (class range_label_for_type_mismatch): New class.
3435 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
3436 of format_warning_va.
3437 (fmtwarn_n): Likewise for new params of format_warning_n_va.
3438 * lto-wrapper.c (merge_and_complain): Add
3439 OPT_fdiagnostics_show_labels to the "pick one setting" options.
3440 (append_compiler_options): Likewise to the dropped options.
3441 (append_diag_options): Likewise to the passed-on options.
3442 * opts.c (common_handle_option): Handle the new option.
3443 * selftest-diagnostic.c
3444 (test_diagnostic_context::test_diagnostic_context): Enable
3446 * substring-locations.c: Include "gcc-rich-location.h".
3447 (format_warning_n_va): Add "fmt_label" and "param_label" params
3448 and use them as appropriate.
3449 (format_warning_va): Add "fmt_label" and "param_label" params,
3450 passing them on to format_warning_n_va.
3451 (format_warning_at_substring): Likewise.
3452 (format_warning_at_substring_n): Likewise.
3453 * substring-locations.h (format_warning_va): Add "fmt_label" and
3454 "param_label" params.
3455 (format_warning_n_va): Likewise.
3456 (format_warning_at_substring): Likewise.
3457 (format_warning_at_substring_n): Likewise.
3458 * toplev.c (general_init): Initialize global_dc->show_labels_p.
3460 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
3463 * builtins.c (expand_builtin_memcmp): Do not expand the call
3464 when overflow is detected.
3466 2018-08-15 Martin Sebor <msebor@redhat.com>
3468 PR tree-optimization/71625
3469 * config/aarch64/aarch64-builtins.c
3470 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3472 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
3474 * config/s390/s390.c (s390_reorg): Remove loop.
3476 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3479 (darwin_function_switched_text_sections): Delete.
3480 * gcc/config/darwin.h
3481 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3483 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3486 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3487 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3488 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3490 2018-08-15 Martin Liska <mliska@suse.cz>
3492 PR tree-optimization/86925
3493 * predict.c (expr_expected_value_1): When taking
3494 later predictor, assign also probability.
3495 Use fold_build2_initializer_loc in order to fold
3496 the expression in -frounding-math.
3498 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3500 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3502 (expand_vec_perm_1): Try the new method.
3504 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3507 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3508 Check whether lra_live_max_point is 0 before dividing.
3510 2018-08-14 Martin Sebor <msebor@redhat.com>
3512 PR tree-optimization/86650
3513 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3514 (vrp_prop::check_mem_ref): Same.
3516 2018-08-13 Liu Hao <lh_mouse@126.com>
3518 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3519 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3520 and clear it thereafter, as it only works for DBCS.
3522 2018-08-13 Liu Hao <lh_mouse@126.com>
3524 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3525 handle returned by _get_osf_handle().
3527 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3529 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3530 for folding vec_perm.
3532 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3534 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3535 Add support for gimple-folding of vec_pack() and vec_unpack()
3538 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3540 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3541 vec_xst variants to the list.
3542 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3543 vector loads and stores.
3545 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3547 * config.gcc (rs6000-ibm-aix4.x): Delete.
3548 (rs6000-ibm-aix5.1): Delete.
3549 (rs6000-ibm-aix5.2): Delete.
3550 (rs6000-ibm-aix5.3): Delete.
3551 * config/rs6000/aix43.h: Delete.
3552 * config/rs6000/aix51.h: Delete.
3553 * config/rs6000/aix52.h: Delete.
3554 * config/rs6000/t-aix43: Delete.
3556 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3558 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3560 (s390_decompose_address): Factor out constant pool ref
3563 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3565 * config/nds32/nds32-predicates.c
3566 (nds32_can_use_bclr_p): Change return type as bool.
3567 (nds32_can_use_bset_p): Ditto.
3568 (nds32_can_use_btgl_p): Ditto.
3569 (nds32_can_use_bitci_p): Ditto.
3570 * config/nds32/nds32-protos.h
3571 (nds32_can_use_bclr_p): Change declaration.
3572 (nds32_can_use_bset_p): Ditto.
3573 (nds32_can_use_btgl_p): Ditto.
3574 (nds32_can_use_bitci_p): Ditto.
3576 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3578 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3579 Support -msched-prolog-epilog option.
3580 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3582 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3584 * common/config/nds32/nds32-common.c
3585 (nds32_option_optimization_table): Enalbe -malways-align.
3587 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3589 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3591 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3592 OPT_misr_secure_ case.
3593 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3594 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3595 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3597 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3598 (nds32_isr_info): New field security_level.
3599 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3600 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3601 * config/nds32/nds32.opt (misr-secure): New option.
3602 * config/nds32/nds32_init.inc: New file.
3603 * config/nds32/nds32_isr.h: New file.
3605 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3607 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3608 Update comment for atomic instructions.
3609 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3610 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3612 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3613 load for __ATOMIC_SEQ_CST model.
3614 (atomic_loaddi_1): Remove float register target.
3615 (atomic_storedi): Handle CONST_INT values.
3616 (atomic_storedi_1): Remove float register source. Add special case
3618 (memory_barrier): New expander and insn.
3620 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3622 PR tree-optimization/86835
3623 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3624 new_stmt after def_gsi, make sure to insert new_square_stmt after
3625 that stmt, not 2 stmts before it.
3627 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3630 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3633 2018-08-10 Martin Liska <mliska@suse.cz>
3636 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3638 * builtins.c (expand_builtin_expect_with_probability):
3640 (expand_builtin_expect_with_probability): New function.
3641 (build_builtin_expect_predicate): Add new argumnet probability
3642 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3643 (fold_builtin_expect):
3646 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3647 * builtins.h (fold_builtin_expect): Set new argument.
3648 * doc/extend.texi: Document __builtin_expect_with_probability.
3649 * doc/invoke.texi: Likewise.
3650 * gimple-fold.c (gimple_fold_call): Pass new argument.
3651 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3652 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3653 * predict.c (get_predictor_value): New function.
3654 (expr_expected_value): Add new argument probability. Assume
3655 that predictor and probability are always non-null.
3656 (expr_expected_value_1): Likewise. For __builtin_expect and
3657 __builtin_expect_with_probability set probability. Handle
3658 combination in binary expressions.
3659 (tree_predict_by_opcode): Simplify code by simply calling
3660 get_predictor_value.
3661 (pass_strip_predict_hints::execute): Add handling of
3662 BUILT_IN_EXPECT_WITH_PROBABILITY.
3663 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3665 * tree.h (DECL_BUILT_IN_P): New function.
3667 2018-08-10 Martin Liska <mliska@suse.cz>
3669 PR tree-optimization/85799
3670 * passes.def: Add argument for pass_strip_predict_hints.
3671 * predict.c (class pass_strip_predict_hints): Add new argument
3673 (strip_predictor_early): New function.
3674 (pass_strip_predict_hints::execute): Call the function to
3676 (strip_predict_hints): New function.
3677 * predict.def: Fix comment.
3679 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
3681 * Makefile.in: Clarify which tm.texi to copy over to assert the
3682 right to grant a GFDL license for all.
3684 2018-08-09 Jeff Law <law@redhat.com>
3686 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3689 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
3691 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3694 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3696 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3697 reductions for variable-length vectors.
3699 2018-08-09 David Malcolm <dmalcolm@redhat.com>
3702 * common.opt (fdiagnostics-show-line-numbers): New option.
3703 * diagnostic-show-locus.c (class layout): Add fields
3704 "m_show_line_numbers_p" and "m_linenum_width";
3705 (num_digits): New function.
3706 (test_num_digits): New function.
3707 (layout::layout): Initialize new fields. Update m_x_offset
3708 logic to handle any left margin.
3709 (layout::print_source_line): Print line number when requested.
3710 (layout::start_annotation_line): New member function.
3711 (layout::print_annotation_line): Call it.
3712 (layout::print_leading_fixits): Likewise.
3713 (layout::print_trailing_fixits): Likewise. Update calls to
3714 move_to_column for new parameter.
3715 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3716 it to potentially call start_annotation_line.
3717 (layout::show_ruler): Call start_annotation_line.
3718 (selftest::test_line_numbers_multiline_range): New selftest.
3719 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3720 and selftest::test_line_numbers_multiline_range.
3721 * diagnostic.c (diagnostic_initialize): Initialize
3722 show_line_numbers_p.
3723 * diagnostic.h (struct diagnostic_context): Add field
3724 "show_line_numbers_p".
3725 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3726 -fno-diagnostics-show-line-numbers.
3727 * dwarf2out.c (gen_producer_string): Add
3728 OPT_fdiagnostics_show_line_numbers to the ignored options.
3729 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3730 one setting" options.
3731 (append_compiler_options): Likewise to the dropped options.
3732 (append_diag_options): Likewise to the passed-on options.
3733 * opts.c (common_handle_option): Handle the new option.
3734 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3736 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
3738 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3739 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
3740 third argument of type "const signed char" to descriptions of
3741 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3742 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3743 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3744 __builtin_bcdsub_ov functions.
3746 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3748 PR tree-optimization/86858
3749 * tree-vect-loop.c (vect_is_simple_reduction): Restore
3750 flow_bb_inside_loop_p calls.
3752 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3754 PR tree-optimization/86871
3755 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3756 instead of gimple_assign_lhs.
3758 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
3761 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3762 register constraint to operand 0.
3763 (add<mode>3_carryinC): Likewise.
3764 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3766 2018-08-09 Martin Liska <mliska@suse.cz>
3769 * common.opt: Remove extra line.
3771 2018-08-09 Martin Liska <mliska@suse.cz>
3773 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3774 at the end of a line, make first letter capital and end up
3775 a sentence with a dot.
3776 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3777 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3778 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3779 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3780 (PARAM_MAX_ISL_OPERATIONS): Likewise.
3781 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3782 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3783 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3784 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3785 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3786 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3787 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3788 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3789 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3790 (PARAM_TREE_REASSOC_WIDTH): Likewise.
3791 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3792 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3793 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3795 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
3798 * config/s390/s390.c (s390_option_override_internal): Reduce the
3799 stack-clash-protection-probe-interval param if it would be too big
3802 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
3805 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3806 * config/m68k/m68k.c (handle_move_double): Don't call
3807 m68k_final_prescan_insn.
3808 (m68k_adjust_decorated_operand): Renamed from
3809 m68k_final_prescan_insn, remove first and third operand and
3811 (print_operand): Call it.
3812 (print_operand_address): Call it.
3814 2018-08-08 Nathan Sidwell <nathan@acm.org>
3816 * diagnostic.c (diagnostic_report_current_module): Use
3817 linemap_included_from & linemap_included_from_linemap.
3819 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3821 * config/aarch64/aarch64-cores.def: Add phecda core.
3822 * config/aarch64/aarch64-tune.md: Regenerate.
3823 * doc/invoke.texi: Add phecda core.
3825 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3828 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3830 * config/s390/s390.md ("movti"): Add more alternatives for
3831 constant to GPR copies.
3833 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3835 * config/s390/s390.c: Fix whitespace damage throughout the file.
3836 * config/s390/s390.h: Likewise.
3837 * config/s390/tpf.h: Likewise.
3839 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3841 * config/s390/s390.c (s390_loadrelative_operand_p):
3842 Remove TARGET_CPU_ZARCH usages.
3843 (s390_rtx_costs): Likewise.
3844 (s390_legitimate_constant_p): Likewise.
3845 (s390_cannot_force_const_mem): Likewise.
3846 (legitimate_reload_constant_p): Likewise.
3847 (s390_preferred_reload_class): Likewise.
3848 (legitimize_pic_address): Likewise.
3849 (legitimize_tls_address): Likewise.
3850 (s390_split_branches): Removed.
3851 (s390_add_execute): Removed.
3852 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3853 (s390_mainpool_start): Likewise.
3854 (s390_mainpool_finish): Likewise.
3855 (s390_mainpool_cancel): Removed.
3856 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3857 (s390_chunkify_cancel): Likewise.
3858 (s390_return_addr_rtx): Likewise.
3859 (s390_register_info): Remove split_branches_pending_p uages.
3860 (s390_optimize_register_info): Likewise.
3861 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3862 split_branches_pending_p usages.
3863 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3864 (s390_load_got): Likewise.
3865 (s390_expand_split_stack_prologue): Likewise.
3866 (output_asm_nops): Likewise.
3867 (s390_function_profiler): Likewise.
3868 (s390_emit_call): Likewise.
3869 (s390_conditional_register_usage): Likewise.
3870 (s390_optimize_prologue): Likewise.
3871 (s390_reorg): Remove TARGET_CPU_ZARCH and
3872 split_branches_pending_p usages.
3873 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3875 (s390_output_indirect_thunk_function): Likewise.
3876 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3877 (TARGET_CPU_ZARCH_P): Removed.
3878 (struct machine_function): Remove split_branches_pending_p.
3879 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3881 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3883 * common/config/s390/s390-common.c (processor_flags_table):
3885 * config.gcc: Remove with_arch/with_tune support.
3886 * config/s390/2064.md: Remove cpu attribute comparisons.
3887 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3889 * config/s390/linux.h (ASM_SPEC):
3890 Remove -march support.
3891 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3892 Use a table to get an arch level.
3893 * config/s390/s390-opts.h (enum processor_type):
3895 * config/s390/s390.c
3896 (processor_table): Remove entries, add arch_level values.
3897 (s390_issue_rate): Remove cases.
3898 (s390_option_override): Adjust
3899 s390_option_override_internal() call.
3900 (s390_option_override_internal): Remove deprecation warning.
3901 (s390_valid_target_attribute_tree): Adjust
3902 s390_option_override_internal() call.
3903 * config/s390/s390.h (struct s390_processor):
3904 Share with s390-c.c, add arch_level field.
3905 * config/s390/s390.md:
3906 Remove occurrences in cpu attribute.
3907 * config/s390/s390.opt: Remove -march/-mtune support.
3908 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3909 * doc/invoke.texi: Remove deprecation warning.
3911 2018-08-08 Luis Machado <luis.machado@linaro.org>
3913 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3915 (qdf24xx_tunings): Set vector cost structure to
3916 qdf24xx_vector_cost.
3918 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3919 <register_sextend>: Set to 3.
3921 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3924 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3925 * config/aarch64/aarch64-simd.md
3926 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3927 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3928 (aarch64_frecpx<mode>): ...this new pattern.
3929 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3930 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3932 2018-08-07 Martin Liska <mliska@suse.cz>
3935 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3936 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3937 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3938 * doc/extend.texi: Document that malloc attribute adds
3941 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3944 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3945 Define to speculation_safe_value_not_needed.
3947 2018-08-06 Jeff Law <law@redhat.com>
3949 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3950 the vr_values instance to cprop_into_stmt.
3951 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3952 (cprop_operand): Also query EVRP to determine if OP is a constant.
3954 2018-08-06 Nathan Sidwell <nathan@acm.org>
3956 * diagnostic.c (diagnostic_report_current_module): Reroll
3957 included-at loop. Translate text.
3959 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3961 * function-tests.c (selftest::test_expansion_to_rtl): Call
3962 free_after_compilation.
3964 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3966 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3968 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3970 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3971 loops with memory block operations from getting unrolled.
3973 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3976 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3977 Define to speculation_safe_value_not_needed.
3979 2018-08-06 Jeff Law <law@redhat.com>
3981 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3984 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3987 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3988 with all enabled __intN types.
3990 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3992 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3994 * alias.c (record_set): Check for clobber high.
3995 * cfgexpand.c (expand_gimple_stmt): Likewise.
3996 * combine-stack-adj.c (single_set_for_csa): Likewise.
3997 * combine.c (find_single_use_1): Likewise.
3998 (set_nonzero_bits_and_sign_copies): Likewise.
3999 (get_combine_src_dest): Likewise.
4000 (is_parallel_of_n_reg_sets): Likewise.
4001 (try_combine): Likewise.
4002 (record_dead_and_set_regs_1): Likewise.
4003 (reg_dead_at_p_1): Likewise.
4004 (reg_dead_at_p): Likewise.
4005 * dce.c (deletable_insn_p): Likewise.
4006 (mark_nonreg_stores_1): Likewise.
4007 (mark_nonreg_stores_2): Likewise.
4008 * df-scan.c (df_find_hard_reg_defs): Likewise.
4009 (df_uses_record): Likewise.
4010 (df_get_call_refs): Likewise.
4011 * dwarf2out.c (mem_loc_descriptor): Likewise.
4012 * haifa-sched.c (haifa_classify_rtx): Likewise.
4013 * ira-build.c (create_insn_allocnos): Likewise.
4014 * ira-costs.c (scan_one_insn): Likewise.
4015 * ira.c (equiv_init_movable_p): Likewise.
4016 (rtx_moveable_p): Likewise.
4017 (interesting_dest_for_shprep): Likewise.
4018 * jump.c (mark_jump_label_1): Likewise.
4019 * postreload-gcse.c (record_opr_changes): Likewise.
4020 * postreload.c (reload_cse_simplify): Likewise.
4021 (struct reg_use): Add source expr.
4022 (reload_combine): Check for clobber high.
4023 (reload_combine_note_use): Likewise.
4024 (reload_cse_move2add): Likewise.
4025 (move2add_note_store): Likewise.
4026 * print-rtl.c (print_pattern): Likewise.
4027 * recog.c (decode_asm_operands): Likewise.
4028 (store_data_bypass_p): Likewise.
4029 (if_test_bypass_p): Likewise.
4030 * regcprop.c (kill_clobbered_value): Likewise.
4031 (kill_set_value): Likewise.
4032 * reginfo.c (reg_scan_mark_refs): Likewise.
4033 * reload1.c (maybe_fix_stack_asms): Likewise.
4034 (eliminate_regs_1): Likewise.
4035 (elimination_effects): Likewise.
4036 (mark_not_eliminable): Likewise.
4037 (scan_paradoxical_subregs): Likewise.
4038 (forget_old_reloads_1): Likewise.
4039 * reorg.c (find_end_label): Likewise.
4040 (try_merge_delay_insns): Likewise.
4041 (redundant_insn): Likewise.
4042 (own_thread_p): Likewise.
4043 (fill_simple_delay_slots): Likewise.
4044 (fill_slots_from_thread): Likewise.
4045 (dbr_schedule): Likewise.
4046 * resource.c (update_live_status): Likewise.
4047 (mark_referenced_resources): Likewise.
4048 (mark_set_resources): Likewise.
4049 * rtl.c (copy_rtx): Likewise.
4050 * rtlanal.c (reg_referenced_p): Likewise.
4051 (single_set_2): Likewise.
4052 (noop_move_p): Likewise.
4053 (note_stores): Likewise.
4054 * sched-deps.c (sched_analyze_reg): Likewise.
4055 (sched_analyze_insn): Likewise.
4057 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4059 * cse.c (invalidate_reg): New function extracted from...
4060 (invalidate): ...here.
4061 (canonicalize_insn): Check for clobber high.
4062 (invalidate_from_clobbers): invalidate clobber highs.
4063 (invalidate_from_sets_and_clobbers): Likewise.
4064 (count_reg_usage): Check for clobber high.
4065 (insn_live_p): Likewise.
4066 * cselib.c (cselib_expand_value_rtx_1):Likewise.
4067 (cselib_invalidate_regno): Check for clobber in setter.
4068 (cselib_invalidate_rtx): Pass through setter.
4069 (cselib_invalidate_rtx_note_stores):
4070 (cselib_process_insn): Check for clobber high.
4071 * cselib.h (cselib_invalidate_rtx): Add operand.
4073 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4075 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
4076 (mark_not_eliminable): Likewise.
4077 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
4078 * lra-lives.c (process_bb_lives): Check for clobber high.
4079 * lra.c (new_insn_reg): Remember clobber highs.
4080 (collect_non_operand_hard_regs): Check for clobber high.
4081 (lra_set_insn_recog_data): Likewise.
4082 (add_regs_to_insn_regno_info): Likewise.
4083 (lra_update_insn_regno_info): Likewise.
4085 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4087 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
4088 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
4090 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4092 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
4093 (copy_insn_1): Likewise.
4094 (gen_hard_reg_clobber_high): New gen function.
4095 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
4096 * genemit.c (gen_exp): Likewise.
4097 (gen_emit_seq): Pass through info.
4098 (gen_insn): Check for CLOBBER_HIGH.
4099 (gen_expand): Pass through info.
4100 (gen_split): Likewise.
4101 (output_add_clobbers): Likewise.
4102 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
4103 (remove_clobbers): Likewise.
4104 * rtl.h (gen_hard_reg_clobber_high): New declaration.
4106 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4108 * doc/rtl.texi (clobber_high): Add.
4109 (parallel): Add in clobber high
4110 * rtl.c (rtl_check_failed_code3): Add function.
4111 * rtl.def (CLOBBER_HIGH): Add expression.
4112 * rtl.h (RTL_CHECKC3): Add macro.
4113 (rtl_check_failed_code3): Add declaration.
4114 (XC3EXP): Add macro.
4116 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
4119 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
4120 cfun->machine->max_used_stack_alignment if needed.
4122 2018-08-04 Martin Sebor <msebor@redhat.com>
4124 PR tree-optimization/86571
4125 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
4128 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
4131 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4134 2018-08-03 Jeff Law <law@redhat.com>
4137 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4138 Define to speculation_safe_value_not_needed.
4140 2018-08-03 David Malcolm <dmalcolm@redhat.com>
4142 * doc/gcov.texi (-x): Remove duplicate "to".
4143 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
4144 (-Wif-not-aligned): Remove duplicate "is".
4145 (-flto): Remove duplicate "the".
4146 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
4147 duplicate "v5.00.b".
4148 (MSP430 Options): Remove duplicate "and" from the description
4149 of "-mgprel-sec=regexp".
4150 (x86 Options): Remove duplicate copies of "vmldLog102" and
4151 vmlsLog104 from description of "-mveclibabi=type".
4153 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
4155 * internal-fn.h (first_commutative_argument): Declare.
4156 * internal-fn.c (first_commutative_argument): New function.
4157 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
4158 restrictions for pattern statements. Use first_commutative_argument
4159 to look for commutative operands in calls to internal functions.
4161 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
4163 * Makefile.in (wide-int-range.o): New.
4164 * tree-vrp.c: Move all the wide_int_* functions to...
4165 * wide-int-range.cc: ...here.
4166 * tree-vrp.h: Move all the wide_int_* prototypes to...
4167 * wide-int-range.h: ...here.
4169 2018-08-03 Tom de Vries <tdevries@suse.de>
4171 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
4173 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
4174 * except.c (output_function_exception_table): Do early exit if
4175 targetm_common.except_unwind_info (&global_options) == UI_NONE.
4177 2018-08-03 Martin Liska <mliska@suse.cz>
4179 * predict.c (dump_prediction): Change to 2 digits
4182 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
4184 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
4186 (falkor_am_1_gtov_gtov): ... a new insn reservation.
4188 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
4190 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
4191 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
4192 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
4193 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
4194 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
4196 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4198 * diagnostic-show-locus.c (diagnostic_show_locus): Use
4199 pp_take_prefix when saving the existing prefix.
4200 * diagnostic.c (diagnostic_append_note): Likewise.
4201 * langhooks.c (lhd_print_error_function): Likewise.
4202 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
4203 param's type. Free the existing prefix.
4204 (pp_take_prefix): New function.
4205 (pretty_printer::pretty_printer): Drop the prefix parameter.
4206 Rename the length parameter to match the comment.
4207 (pretty_printer::~pretty_printer): Free the prefix.
4208 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
4210 (struct pretty_printer): Drop the "const" from "prefix" field's
4211 type and clarify memory management.
4212 (pp_set_prefix): Drop the "const" from the 2nd param.
4213 (pp_take_prefix): New decl.
4215 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4217 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
4218 for word_mode_ok here instead of passing as argument.
4219 (expand_block_compare): Change select_block_compare_mode() call.
4220 (expand_strncmp_gpr_sequence): New function.
4221 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
4223 2018-08-02 Jeff Law <law@redhat.com>
4226 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4227 Define to speculation_safe_value_not_needed.
4230 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4231 Define to speculation_safe_value_not_needed.
4233 2018-08-02 Tom de Vries <tdevries@suse.de>
4236 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
4237 function. Return UI_TARGET unconditionally.
4238 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
4239 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
4241 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4243 * genemit.c (print_overload_test): Fix typo.
4245 2018-08-02 Richard Biener <rguenther@suse.de>
4247 PR tree-optimization/86816
4248 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
4249 which checks for value availability before querying it.
4250 (gvn_uses_equal): Use it.
4251 (same_succ_hash): Likewise.
4252 (gimple_equal_p): Likewise.
4254 2018-08-02 Nick Clifton <nickc@redhat.com>
4257 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4258 Define to speculation_safe_value_not_needed.
4261 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4262 Define to speculation_safe_value_not_needed.
4265 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4266 Define to speculation_safe_value_not_needed.
4269 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4270 Define to speculation_safe_value_not_needed.
4273 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4274 Define to speculation_safe_value_not_needed.
4277 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4278 Define to speculation_safe_value_not_needed.
4281 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4282 Define to speculation_safe_value_not_needed.
4285 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4286 Define to speculation_safe_value_not_needed.
4289 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4290 speculation_safe_value_not_needed.
4293 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
4294 to speculation_safe_value_not_needed.
4296 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4298 * doc/md.texi: Expand the documentation of instruction names
4299 to mention port-local uses. Document '@' in pattern names.
4300 * read-md.h (overloaded_instance, overloaded_name): New structs.
4302 (md_reader::handle_overloaded_name): New member function.
4303 (md_reader::get_overloads): Likewise.
4304 (md_reader::m_first_overload): New member variable.
4305 (md_reader::m_next_overload_ptr): Likewise.
4306 (md_reader::m_overloads_htab): Likewise.
4307 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
4308 m_next_overload_ptr and m_overloads_htab.
4309 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
4310 (get_mode_token, get_code_token, get_int_token): New functions.
4311 (map_attr_string): Add an optional argument that passes back
4312 the associated iterator.
4313 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
4314 (md_reader::handle_overloaded_name, add_overload_instance): New
4316 (apply_iterators): Handle '@' names. Report an error if '@'
4317 is used without iterators.
4318 (initialize_iterators): Initialize the new iterator_group fields.
4319 * genopinit.c (handle_overloaded_code_for)
4320 (handle_overloaded_gen): New functions.
4321 (main): Use them to print declarations of maybe_code_for_* and
4322 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
4323 * genemit.c (print_overload_arguments, print_overload_test)
4324 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
4325 (main): Use it to print definitions of maybe_code_for_* and
4326 maybe_gen_* functions.
4327 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
4328 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
4329 instead of explicit mode checks.
4330 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
4331 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
4332 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
4333 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
4334 (aarch64_expand_compare_and_swap): Likewise
4335 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
4336 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
4337 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
4338 (aarch64_constant_pool_reload_icode): Delete.
4339 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
4340 instead of aarch64_constant_pool_reload_icode. Use
4341 code_for_aarch64_reload_mov instead of explicit mode checks.
4342 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
4343 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
4344 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
4345 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
4346 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
4347 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
4348 (aarch64_atomic_load_op_code): Delete.
4349 (aarch64_emit_atomic_load_op): Likewise.
4350 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
4351 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
4352 instead of aarch64_emit_atomic_load_op.
4353 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
4354 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
4355 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
4356 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
4357 character before the pattern name.
4358 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
4359 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
4360 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
4361 (aarch64_frecps<mode>): Likewise.
4362 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
4363 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
4364 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
4365 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
4366 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
4368 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4370 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
4371 Allow HFmode constants if TARGET_FP_F16INST.
4373 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
4376 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
4377 No longer check last store for clobber of address register.
4379 2018-08-02 Martin Liska <mliska@suse.cz>
4381 PR gcov-profile/86817
4382 * gcov.c (process_all_functions): New function.
4384 (process_file): Move functions processing to
4385 process_all_functions.
4387 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4389 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
4390 "const" to the "gimple *" and "rtx_insn *" parameters.
4391 * dumpfile.h (dump_user_location_t::dump_user_location_t):
4393 (dump_location_t::dump_location_t): Likewise.
4395 2018-08-01 Martin Sebor <msebor@redhat.com>
4397 PR tree-optimization/86650
4398 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
4399 rather than a "gcall *". Directly pass the data of interest
4400 to percent_K_format, rather than building a temporary CALL_EXPR
4402 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
4403 (gimple_fold_builtin_strncat): Adjust.
4404 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
4405 gcall* argument with gimple*.
4406 * gimple-ssa-warn-restrict.c (check_call): Same.
4407 (wrestrict_dom_walker::before_dom_children): Same.
4408 (builtin_access::builtin_access): Same.
4409 (check_bounds_or_overlap): Same
4410 (maybe_diag_overlap): Same.
4411 (maybe_diag_offset_bounds): Same.
4412 * tree-diagnostic.c (default_tree_printer): Move usage of
4413 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
4415 * tree-pretty-print.c (percent_K_format): Add argument.
4416 * tree-pretty-print.h: Add argument.
4417 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
4418 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
4419 (maybe_diag_stxncpy_trunc): Same.
4420 (handle_builtin_stxncpy): Same.
4421 (handle_builtin_strcat): Same.
4423 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4425 * match.pd: Optimise pointer range checks.
4427 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4429 PR tree-optimization/86758
4430 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
4431 to remove pattern statements.
4433 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4435 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
4436 result of dfs_enumerate_from when constructing stmt_vec_infos,
4437 instead of additionally calling get_loop_body.
4439 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4441 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4443 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4444 When creating an iv, assert that the step is not known to be zero.
4445 (vect_setup_realignment): Update call accordingly.
4446 * tree-vect-stmts.c (vectorizable_store): Likewise.
4447 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
4449 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4451 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4452 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4453 (vectorizable_reduction): Likewise.
4454 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4455 (vect_detect_hybrid_slp_stmts): Likewise.
4456 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4458 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
4460 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4461 (wide_int_set_zero_nonzero_bits): ...this.
4462 (zero_nonzero_bits_from_vr): Rename to...
4463 (vrp_set_zero_nonzero_bits): ...this.
4464 (extract_range_from_multiplicative_op_1): Abstract wide int
4466 (wide_int_range_multiplicative_op): ...here.
4467 (extract_range_from_binary_expr_1): Extract wide int binary
4468 operations into their own functions.
4469 (wide_int_range_lshift): New.
4470 (wide_int_range_can_optimize_bit_op): New.
4471 (wide_int_range_shift_undefined_p): New.
4472 (wide_int_range_bit_xor): New.
4473 (wide_int_range_bit_ior): New.
4474 (wide_int_range_bit_and): New.
4475 (wide_int_range_trunc_mod): New.
4476 (extract_range_into_wide_ints): New.
4477 (vrp_shift_undefined_p): New.
4478 (extract_range_from_multiplicative_op): New.
4479 (vrp_can_optimize_bit_op): New.
4480 * tree-vrp.h (value_range::dump): New.
4481 (wide_int_range_multiplicative_op): New.
4482 (wide_int_range_lshift):New.
4483 (wide_int_range_shift_undefined_p): New.
4484 (wide_int_range_bit_xor): New.
4485 (wide_int_range_bit_ior): New.
4486 (wide_int_range_bit_and): New.
4487 (wide_int_range_trunc_mod): New.
4488 (zero_nonzero_bits_from_bounds): Rename to...
4489 (wide_int_set_zero_nonzero_bits): ...this.
4490 (zero_nonzero_bits_from_vr): Rename to...
4491 (vrp_set_zero_nonzero_bits): ...this.
4492 (range_easy_mask_min_max): Rename to...
4493 (wide_int_range_can_optimize_bit_op): this.
4494 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4495 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4497 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4499 * tree-vectorizer.h (vect_orig_stmt): New function.
4500 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4501 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4502 (vect_create_epilog_for_reduction): Likewise.
4503 (vectorizable_live_operation): Likewise.
4504 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4505 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4506 * tree-vect-stmts.c (vectorizable_call): Likewise.
4507 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4509 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4511 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4513 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4514 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4515 (vect_transform_loop): Likewise.
4516 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4518 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4520 * tree-vectorizer.h (vect_schedule_slp): Return void.
4521 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4522 (vect_schedule_slp): Likewise.
4524 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4526 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4528 (vect_transform_loop): Update calls accordingly. Schedule SLP
4529 instances before the main loop, if any exist.
4531 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4533 PR tree-optimization/86749
4534 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4535 If the lhs is used in a COND_EXPR, check that it is being used
4536 as the "then" or "else" value.
4538 2018-08-01 Tom de Vries <tdevries@suse.de>
4541 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4542 speculation_safe_value_not_needed.
4544 2018-08-01 Richard Biener <rguenther@suse.de>
4546 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4549 2018-08-01 Martin Liska <mliska@suse.cz>
4551 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4552 format how successful transformation is dumped.
4553 (gimple_mod_pow2_value_transform): Likewise.
4554 (gimple_mod_subtract_transform): Likewise.
4555 (gimple_stringops_transform): Likewise.
4557 2018-08-01 Martin Liska <mliska@suse.cz>
4560 * value-prof.c (interesting_stringop_to_profile_p):
4561 Simplify the code and add BUILT_IN_MEMMOVE.
4562 (gimple_stringops_transform): Likewise.
4564 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4566 * config/aarch64/aarch64-simd.md
4567 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4568 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4569 use GPI iterator instead of SI mode.
4571 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4573 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4574 rs6000_speculation_barrier.
4575 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4576 new barrier pattern name.
4578 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4580 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4581 (speculation_barrier): New insn.
4583 2018-08-01 Richard Biener <rguenther@suse.de>
4586 * graphite.h: Include isl/id.h and isl/space.h to allow build
4589 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4592 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4593 mode for COFF targets.
4594 * defaults.h (TARGET_COFF): Define.
4595 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4596 TARGET_COFF): Define.
4597 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4598 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4599 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4601 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4603 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4604 * gimple-streamer-out.c (output_bb): Save it.
4605 * lto-streamer-in.c (input_struct_function_base): Restore
4606 instance discriminator if available. Create map on demand.
4607 * lto-streamer-out.c (output_struct_function_base): Save it if
4609 * final.c (decl_to_instance_map): Document LTO strategy.
4611 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4612 Olivier Hainque <hainque@adacore.com>
4614 * debug.h (decl_to_instance_map_t): New type.
4615 (decl_to_instance_map): Declare.
4616 (maybe_create_decl_to_instance_map): New inline function.
4617 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4618 to track basic block discriminators.
4619 (final_start_function_1): Initialize them.
4620 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4622 (decl_to_instance_map): New variable.
4623 (map_decl_to_instance, maybe_set_discriminator): New functions.
4624 (notice_source_line): Set discriminator.
4626 2018-07-31 Ian Lance Taylor <iant@golang.org>
4628 * targhooks.c (default_have_speculation_safe_value): Add
4631 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4633 * dump-context.h: Include "pretty-print.h".
4634 (dump_context::refresh_dumps_are_enabled): New decl.
4635 (dump_context::emit_item): New decl.
4636 (class dump_context): Add fields "m_test_pp" and
4638 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4639 (temp_dump_context::get_dumped_text): New decl.
4640 (class temp_dump_context): Add field "m_pp".
4641 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4642 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4644 (set_dump_file): Update for above change.
4645 (set_alt_dump_file): Likewise.
4646 (dump_loc): New overload, taking a pretty_printer *.
4647 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4648 to any test pretty-printer.
4649 (make_item_for_dump_gimple_stmt): New function, adapted from
4650 optinfo::add_gimple_stmt.
4651 (dump_context::dump_gimple_stmt): Call it, and use the result,
4652 eliminating the direct usage of dump_file and alt_dump_file in
4653 favor of indirectly using them via emit_item.
4654 (make_item_for_dump_gimple_expr): New function, adapted from
4655 optinfo::add_gimple_expr.
4656 (dump_context::dump_gimple_expr): Call it, and use the result,
4657 eliminating the direct usage of dump_file and alt_dump_file in
4658 favor of indirectly using them via emit_item.
4659 (make_item_for_dump_generic_expr): New function, adapted from
4661 (dump_context::dump_generic_expr): Call it, and use the result,
4662 eliminating the direct usage of dump_file and alt_dump_file in
4663 favor of indirectly using them via emit_item.
4664 (make_item_for_dump_printf_va): New function, adapted from
4665 optinfo::add_printf_va.
4666 (make_item_for_dump_printf): New function.
4667 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4668 and use the result, eliminating the direct usage of dump_file and
4669 alt_dump_file in favor of indirectly using them via emit_item.
4670 (make_item_for_dump_dec): New function.
4671 (dump_context::dump_dec): Call it, and use the result,
4672 eliminating the direct usage of dump_file and alt_dump_file in
4673 favor of indirectly using them via emit_item.
4674 (make_item_for_dump_symtab_node): New function, adapted from
4675 optinfo::add_symtab_node.
4676 (dump_context::dump_symtab_node): Call it, and use the result,
4677 eliminating the direct usage of dump_file and alt_dump_file in
4678 favor of indirectly using them via emit_item.
4679 (dump_context::begin_scope): Reimplement, avoiding direct usage
4680 of dump_file and alt_dump_file in favor of indirectly using them
4682 (dump_context::emit_item): New member function.
4683 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4684 Set up test pretty-printer on the underlying context. Call
4685 refresh_dumps_are_enabled.
4686 (temp_dump_context::~temp_dump_context): Call
4687 refresh_dumps_are_enabled.
4688 (temp_dump_context::get_dumped_text): New member function.
4689 (selftest::verify_dumped_text): New function.
4690 (ASSERT_DUMPED_TEXT_EQ): New macro.
4691 (selftest::test_capture_of_dump_calls): Run all tests twice, with
4692 and then without optinfo enabled. Add uses of
4693 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
4694 * dumpfile.h: Update comment for the dump_* API.
4695 * optinfo-emit-json.cc
4696 (selftest::test_building_json_from_dump_calls): Update for new
4697 param for temp_dump_context ctor.
4698 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4699 and "m_owned" field.
4700 (optinfo_item::~optinfo_item): Likewise.
4701 (optinfo::add_item): New member function.
4702 (optinfo::emit): Update comment.
4703 (optinfo::add_string): Delete.
4704 (optinfo::add_printf): Delete.
4705 (optinfo::add_printf_va): Delete.
4706 (optinfo::add_gimple_stmt): Delete.
4707 (optinfo::add_gimple_expr): Delete.
4708 (optinfo::add_tree): Delete.
4709 (optinfo::add_symtab_node): Delete.
4710 (optinfo::add_dec): Delete.
4711 * optinfo.h (class dump_context): New forward decl.
4712 (optinfo::add_item): New decl.
4713 (optinfo::add_string): Delete.
4714 (optinfo::add_printf): Delete.
4715 (optinfo::add_printf_va): Delete.
4716 (optinfo::add_gimple_stmt): Delete.
4717 (optinfo::add_gimple_expr): Delete.
4718 (optinfo::add_tree): Delete.
4719 (optinfo::add_symtab_node): Delete.
4720 (optinfo::add_dec): Delete.
4721 (optinfo::add_poly_int): Delete.
4722 (optinfo_item::optinfo_item): Remove "owned" param.
4723 (class optinfo_item): Remove field "m_owned".
4725 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4728 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4729 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4730 requested variable alignment.
4731 (expand_one_ssa_partition): Likewise.
4732 (expand_one_var): Likewise.
4734 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4736 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4737 to speculation_safe_value_not_needed.
4739 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4741 * targhooks.h (speculation_safe_value_not_needed): New prototype.
4742 * targhooks.c (speculation_safe_value_not_needed): New function.
4743 * target.def (have_speculation_safe_value): Update documentation.
4744 * doc/tm.texi: Regenerated.
4746 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4748 * config/aarch64/iterators.md (ALLI_TI): New iterator.
4749 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4751 (despeculate_copy<ALLI:mode>_insn): New insn.
4752 (despeculate_copyti_insn): New insn.
4753 (despeculate_simple<ALLI:mode>): New insn
4754 (despeculate_simpleti): New insn.
4755 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4757 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4758 aarch64_speculation_safe_value.
4759 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4761 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4763 * config/aarch64/aarch64-speculation.cc: New file.
4764 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4765 before pass_reorder_blocks.
4766 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4768 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4769 X14 and X15 when tracking speculation.
4770 * config/aarch64/aarch64.md (register name constants): Add
4771 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4772 (unspec): Add UNSPEC_SPECULATION_TRACKER.
4773 (speculation_barrier): New insn attribute.
4774 (cmp<mode>): Allow SP in comparisons.
4775 (speculation_tracker): New insn.
4776 (speculation_barrier): Add speculation_barrier attribute.
4777 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4778 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4779 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4781 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4783 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4784 aarch64_track_speculation is true.
4785 (tb<optab><mode>1): Likewise.
4786 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4787 generate CB[N]Z when tracking speculation.
4788 (aarch64_split_compare_and_swap): Likewise.
4789 (aarch64_split_atomic_op): Likewise.
4791 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4793 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4795 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4797 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4798 (speculation_barrier): New insn.
4800 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4802 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4803 * config/arm/arm.md (speculation_barrier): New expand.
4804 (speculation_barrier_insn): New pattern.
4806 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4808 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4809 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4810 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4811 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4813 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4814 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4815 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4816 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4817 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4818 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4819 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4820 * builtins.c (expand_speculation_safe_value): New function.
4821 (expand_builtin): Call it.
4822 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4823 * doc/extend.texi: Document __builtin_speculation_safe_value.
4824 * doc/md.texi: Document "speculation_barrier" pattern.
4825 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4826 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4827 * doc/tm.texi: Regenerated.
4828 * target.def (have_speculation_safe_value, speculation_safe_value): New
4830 * targhooks.c (default_have_speculation_safe_value): New function.
4831 (default_speculation_safe_value): New function.
4832 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4833 (default_speculation_safe_value): Add prototype.
4835 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4837 * dump-context.h (dump_context::dump_loc): New decl.
4838 * dumpfile.c (dump_context::dump_loc): New member function.
4839 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4840 and dump_gimple_stmt.
4841 (dump_context::dump_gimple_expr_loc): Likewise, using
4843 (dump_context::dump_generic_expr_loc): Likewise, using
4845 (dump_context::dump_printf_loc_va): Likewise, using
4847 (dump_context::begin_scope): Explicitly using the global function
4848 "dump_loc", rather than the member function.
4850 2018-07-31 Martin Sebor <msebor@redhat.com>
4852 PR tree-optimization/86741
4853 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4855 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4857 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4858 depend on whether prefetch instructions will be emitted or not.
4859 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4860 will be emitted or not.
4861 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4862 (TARGET_SETMEM_PFD): New macros.
4864 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4866 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4867 (NULL_STMT_VEC_INFO): Delete.
4868 (stmt_vec_info::operator*): Likewise.
4869 (stmt_vec_info::operator gimple *): Likewise.
4870 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4871 of NULL_STMT_VEC_INFO.
4872 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4873 (vect_reassociating_reduction_p): Likewise.
4874 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4875 (vectorizable_store): Likewise.
4876 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4877 (vec_info::free_stmt_vec_infos): Likewise.
4879 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4881 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4882 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4884 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4886 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4887 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4888 (vec_info::free_stmt_vec_info): New private member functions.
4889 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4890 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4891 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4892 set_stmt_vec_info_vec and free_stmt_vec_infos.
4893 * tree-vect-loop.c (_loop_vec_info): Remove call to
4894 set_stmt_vec_info_vec.
4895 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4896 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4897 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4898 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4899 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4900 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4901 with the clearing in the xcalloc.
4902 (stmt_vec_info_vec): Delete.
4903 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4904 (vectorize_loops): Likewise.
4905 (vec_info::~vec_info): Remove argument from call to
4906 free_stmt_vec_infos.
4907 (vec_info::add_stmt): Remove vinfo argument from call to
4910 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4912 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4913 rather than a gimple stmt.
4914 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4915 information for pattern statements when passed the original
4916 statement; instead wait to be passed the pattern statement itself.
4917 Don't call set_vinfo_for_stmt here.
4918 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4919 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4920 stmt_vec_infos here.
4921 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4922 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4923 stmt_vec_infos entry.
4925 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4927 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4928 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4929 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4930 * tree-vect-stmts.c (vectorizable_call): Likewise.
4931 (vectorizable_simd_clone_call): Likewise.
4933 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4935 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4936 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4937 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4938 * tree-vect-loop.c (vect_transform_loop): Likewise.
4939 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4940 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4942 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4944 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4945 (vect_dr_stmt): Delete.
4946 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4947 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4949 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4950 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4951 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4952 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4953 (vect_analyze_data_refs): Likewise.
4954 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4956 (vect_find_same_alignment_drs): Likewise.
4957 (vect_slp_analyze_node_dependences): Update calls accordingly.
4958 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4959 instead of DR_VECT_AUX.
4960 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4961 of a vector data references. Use vec_info::lookup_dr instead of
4963 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4964 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4965 instead of DR_VECT_AUX.
4967 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4969 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4971 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4973 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4974 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4975 (vect_gen_prolog_loop_niters): Likewise.
4977 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4979 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4980 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4981 (vect_known_alignment_in_bytes, vect_dr_behavior)
4982 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4983 instead of data_references. Update calls to other routines for
4984 which the same change has been made.
4985 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4986 dr_vec_infos instead of stmt_vec_infos.
4987 (vect_analyze_data_ref_dependence): Update call accordingly.
4988 (vect_slp_analyze_data_ref_dependence)
4989 (vect_record_base_alignments): Use DR_VECT_AUX.
4990 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4991 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4992 (vector_alignment_reachable_p, vect_get_data_access_cost)
4993 (vect_peeling_supportable, vect_analyze_group_access_1)
4994 (vect_analyze_group_access, vect_analyze_data_ref_access)
4995 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4996 (vect_compile_time_alias, vect_small_gap_p)
4997 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4998 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4999 instead of data_references. Update calls to other routines for
5000 which the same change has been made.
5001 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
5002 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5003 (vect_slp_analyze_and_verify_node_alignment)
5004 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5005 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5006 (vect_setup_realignment): Use dr_vec_infos. Update calls after
5008 (_vect_peel_info::dr): Replace with...
5009 (_vect_peel_info::dr_info): ...this new field.
5010 (vect_peeling_hash_get_most_frequent)
5011 (vect_peeling_hash_choose_best_peeling): Update accordingly.
5012 (vect_peeling_hash_get_lowest_cost):
5013 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
5014 routines for which the same change has been made.
5015 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
5017 * tree-vect-loop-manip.c (get_misalign_in_elems)
5018 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
5020 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5021 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
5022 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
5023 (get_group_load_store_type, get_negative_load_store_type)
5024 (vect_get_data_ptr_increment, vectorizable_store)
5025 (vectorizable_load): Likewise.
5026 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
5027 Update calls to other routines for which the same change has been made.
5029 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5031 * tree-vectorizer.h (vec_info::move_dr): New member function.
5032 (dataref_aux): Rename to...
5033 (dr_vec_info): ...this and add "dr" and "stmt" fields.
5034 (_stmt_vec_info::dr_aux): Update accordingly.
5035 (_stmt_vec_info::data_ref_info): Delete.
5036 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
5037 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
5038 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
5039 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
5041 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
5042 (STMT_VINFO_DR_INFO): New macro.
5043 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
5044 (set_dr_misalignment): Update after rename of dataref_aux.
5045 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
5046 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
5047 initialization of STMT_VINFO_DATA_REF.
5048 * tree-vectorizer.c (vec_info::move_dr): New function.
5049 * tree-vect-patterns.c (vect_recog_bool_pattern)
5050 (vect_recog_mask_conversion_pattern)
5051 (vect_recog_gather_scatter_pattern): Use it.
5052 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
5053 the "dr" and "stmt" fields of dr_vec_info instead of
5054 STMT_VINFO_DATA_REF.
5056 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5058 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
5059 (is_pattern_stmt_p): Use it.
5060 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
5061 on pattern statements.
5063 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5065 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
5066 original stmt as a stmt_vec_info rather than a gimple stmt.
5067 (vect_pattern_recog_1): Take the statement directly as a
5068 stmt_vec_info, rather than via a gimple_stmt_iterator.
5069 Update call to vect_mark_pattern_stmts.
5070 (vect_pattern_recog): Update calls accordingly.
5072 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5074 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
5075 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
5076 a vect_def_type for the first argument.
5077 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
5078 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
5079 operand if it isn't defined by a vectorized statement.
5080 (vect_build_gather_load_calls): Remove the mask_dt argument and
5081 update calls to vect_get_vec_def_for_stmt_copy.
5082 (vectorizable_bswap): Likewise the dt argument.
5083 (vectorizable_call): Update calls to vectorizable_bswap and
5084 vect_get_vec_def_for_stmt_copy.
5085 (vectorizable_simd_clone_call, vectorizable_assignment)
5086 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
5087 (vectorizable_comparison): Update calls to
5088 vect_get_vec_def_for_stmt_copy.
5089 (vectorizable_store): Likewise. Remove now-unnecessary calls to
5091 (vect_get_loop_based_defs): Remove dt argument and update call
5092 to vect_get_vec_def_for_stmt_copy.
5093 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
5094 and vect_get_vec_def_for_stmt_copy.
5095 (vectorizable_load): Update calls to vect_build_gather_load_calls
5096 and vect_get_vec_def_for_stmt_copy.
5097 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5098 (vectorizable_reduction, vectorizable_live_operation): Update calls
5099 to vect_get_vec_def_for_stmt_copy.
5101 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5103 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
5104 and gimple stmt arguments with a stmt_vec_info.
5105 (vect_record_base_alignments): Update calls accordingly.
5106 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
5107 and gimple stmt arguments with a stmt_vec_info.
5108 (vect_build_slp_tree_1): Remove vinfo argument and update call
5109 to vect_record_max_nunits.
5110 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
5111 and vect_record_max_nunits.
5113 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5115 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
5116 file and take a stmt_vec_info instead of a gimple stmt.
5117 (supportable_widening_operation, vect_finish_replace_stmt)
5118 (vect_finish_stmt_generation, vect_get_store_rhs)
5119 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
5120 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
5121 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
5122 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
5123 (vect_create_data_ref_ptr, bump_vector_ptr)
5124 (vect_permute_store_chain, vect_setup_realignment)
5125 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
5126 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
5127 (vectorizable_reduction, vectorizable_induction)
5128 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
5129 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
5130 than gimple stmts as arguments.
5131 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
5132 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
5133 (can_group_stmts_p, vect_check_gather_scatter)
5134 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5135 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
5136 (vect_permute_load_chain, vect_shift_permute_load_chain)
5137 (vect_transform_grouped_load)
5138 (vect_record_grouped_load_vectors): Likewise.
5139 * tree-vect-loop.c (vect_fixup_reduc_chain)
5140 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
5141 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
5142 (vectorizable_reduction, vectorizable_induction)
5143 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
5144 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
5145 (vect_get_load_store_mask): Likewise.
5146 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
5147 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
5148 * tree-vect-stmts.c (vect_mark_relevant)
5149 (is_simple_and_all_uses_invariant)
5150 (exist_non_indexing_operands_for_use_p, process_use)
5151 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
5152 (vect_get_vec_def_for_operand, vect_get_vec_defs)
5153 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
5154 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
5155 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
5156 (get_negative_load_store_type, get_load_store_type)
5157 (vect_check_load_store_mask, vect_check_store_rhs)
5158 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
5159 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5160 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5161 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5162 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
5163 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
5164 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
5165 (supportable_widening_operation): Likewise.
5167 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5169 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
5170 a stmt_vec_info instead of a gcall.
5171 (vect_check_gather_scatter): Update call accordingly.
5172 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
5174 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
5175 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
5176 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
5177 instead of a gimple stmt.
5178 (vect_transform_loop): Update calls accordingly.
5179 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
5180 stmt_vec_infos instead of gimple stmts.
5181 (vect_analyze_slp_instance): Update use accordingly.
5182 * tree-vect-stmts.c (read_vector_array, write_vector_array)
5183 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
5184 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
5185 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
5186 (vect_gen_widened_results_half, vect_get_loop_based_defs)
5187 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
5188 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
5189 down to subroutines.
5191 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5193 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
5194 of the worklist from a vector of gimple stmts to a vector of
5196 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5197 (vect_mark_stmts_to_be_vectorized): Likewise
5199 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5201 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
5202 statement before passing it to vect_analyze_stmt.
5203 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
5204 the chain of phi vector definitions. Track the exit phi via its
5206 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
5207 STMT_VINFO_REDUC_DEF.
5208 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
5209 stmt_vec_infos to handle the statement chains.
5210 (vect_get_slp_defs): Record the first statement in the node
5211 using a stmt_vec_info.
5212 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
5213 statements here and pass their stmt_vec_info down to subroutines.
5214 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
5215 down to vect_finish_stmt_generation.
5216 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
5217 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
5218 stmt_vec_infos to subroutines.
5219 (vect_remove_stores): Use stmt_vec_infos to handle the statement
5222 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5224 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
5225 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
5226 (vect_permute_store_chain, vect_setup_realignment)
5227 (vect_permute_load_chain, vect_shift_permute_load_chain)
5228 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
5229 stmts internally, and when passing values to other vectorizer routines.
5230 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5231 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
5232 (vect_analyze_loop_operations, get_initial_def_for_reduction)
5233 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
5234 (vectorizable_reduction, vectorizable_induction)
5235 (vectorizable_live_operation, vect_transform_loop_stmt)
5236 (vect_transform_loop): Likewise.
5237 * tree-vect-patterns.c (vect_reassociating_reduction_p)
5238 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
5239 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
5240 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5241 (vect_slp_analyze_node_operations_1): Likewise.
5242 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5243 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
5244 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
5245 (vect_finish_stmt_generation_1, get_group_load_store_type)
5246 (get_load_store_type, vect_build_gather_load_calls)
5247 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5248 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5249 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5250 (vectorizable_store, vectorizable_load, vectorizable_condition)
5251 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
5252 (supportable_widening_operation): Likewise.
5253 (vect_get_vector_types_for_stmt): Likewise.
5254 * tree-vectorizer.h (vect_dr_behavior): Likewise.
5256 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5258 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5259 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
5260 (vect_permute_store_chain, vect_permute_load_chain)
5261 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
5262 repeated stmt_vec_info lookups.
5263 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5264 (vect_update_ivs_after_vectorizer): Likewise.
5265 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
5266 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5267 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
5268 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5269 (vect_bb_slp_scalar_cost): Likewise.
5270 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
5272 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5274 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
5275 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
5276 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
5277 to get gassigns and gcalls, rather than operating on generc gimple
5279 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
5280 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
5281 (vectorizable_load, vect_analyze_stmt): Likewise.
5282 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
5284 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5286 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
5287 return stmt_vec_infos rather than gimple stmts. Do not accept
5289 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
5291 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5292 Update use of get_later_stmt.
5293 (vect_get_constant_vectors): Update call accordingly.
5294 (vect_schedule_slp_instance): Likewise
5295 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
5296 (vect_slp_analyze_instance_dependence): Likewise.
5297 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
5299 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5301 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
5302 (stmt_info_for_cost::stmt_info): ...this new field.
5303 (add_stmt_costs): Update accordingly.
5304 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5305 (vect_get_known_peeling_cost): Likewise.
5306 (vect_estimate_min_profitable_iters): Likewise.
5307 * tree-vect-stmts.c (record_stmt_cost): Likewise.
5309 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5311 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
5312 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5313 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5315 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
5317 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5319 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
5320 a gimple stmt to a stmt_vec_info.
5321 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5323 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5325 * tree-vectorizer.h (vec_info::grouped_stores): Change from
5326 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5327 (_loop_vec_info::reduction_chains): Likewise.
5328 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
5330 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5332 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5334 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
5335 a gimple stmt to a stmt_vec_info.
5336 (_stmt_vec_info::next_element): Likewise.
5337 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
5338 (vect_slp_analyze_and_verify_node_alignment)
5339 (vect_analyze_group_access_1, vect_analyze_group_access)
5340 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
5341 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
5342 (vect_supportable_dr_alignment): Update accordingly.
5343 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
5344 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
5345 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
5346 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5347 * tree-vect-slp.c (vect_build_slp_tree_1)
5348 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
5349 (vect_split_slp_store_group, vect_analyze_slp_instance)
5350 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
5351 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
5352 (get_group_load_store_type, get_load_store_type)
5353 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
5354 (vect_transform_stmt, vect_remove_stores): Likewise.
5356 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5358 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5360 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5361 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
5362 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
5363 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
5364 (vector_alignment_reachable_p, vect_get_data_access_cost)
5365 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
5366 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
5367 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5368 (vect_analyze_group_access_1, vect_analyze_group_access)
5369 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
5370 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
5371 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
5372 result of vect_dr_stmt and use the stmt_vec_info instead of
5373 the associated gimple stmt.
5374 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5375 (vect_gen_prolog_loop_niters): Likewise.
5376 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5378 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5380 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
5381 to a vec<stmt_vec_info>.
5382 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
5383 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
5385 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
5386 to a vec<stmt_vec_info>.
5387 (bst_traits::value_type, bst_traits::value_type): Likewise.
5388 (bst_traits::hash): Update accordingly.
5389 (vect_get_and_check_slp_defs): Change the stmts parameter from
5390 a vec<gimple *> to a vec<stmt_vec_info>.
5391 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
5392 (vect_build_slp_tree): Likewise.
5393 (vect_build_slp_tree_2): Likewise. Update uses of
5394 SLP_TREE_SCALAR_STMTS.
5395 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
5396 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
5397 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
5398 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
5399 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
5400 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
5401 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
5402 (vect_get_constant_vectors, vect_get_slp_defs)
5403 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
5404 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
5405 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
5406 instead of gimple stmts.
5407 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
5408 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
5409 (vect_slp_analyze_instance_dependence): Update uses of
5410 SLP_TREE_SCALAR_STMTS.
5411 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5412 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
5413 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
5414 (get_initial_defs_for_reduction): Likewise.
5415 (vect_create_epilog_for_reduction): Likewise.
5416 (vectorize_fold_left_reduction): Likewise.
5417 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
5418 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
5419 (can_vectorize_live_stmts): Likewise.
5421 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5423 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
5424 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5425 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
5427 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
5429 (gather_scalar_reductions): Update after above interface changes.
5430 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
5431 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
5433 (vect_force_simple_reduction): Likewise.
5434 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
5435 LOOP_VINFO_REDUCTIONS.
5436 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5438 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5440 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
5441 a gimple stmt to a stmt_vec_info.
5442 * tree-vect-loop.c (vect_active_double_reduction_p)
5443 (vect_force_simple_reduction, vectorizable_reduction): Update
5446 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5448 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5449 vec<gimple *> to a vec<stmt_vec_info>.
5450 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5451 the reduction_phis argument from a vec<gimple *> to a
5453 (vectorizable_reduction): Likewise the phis local variable that
5454 is passed to vect_create_epilog_for_reduction. Update for new type
5455 of SLP_TREE_VEC_STMTS.
5456 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5457 (vectorizable_live_operation): Likewise.
5458 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5459 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5461 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5463 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5464 a gimple stmt to a stmt_vec_info.
5465 (vectorizable_condition, vectorizable_live_operation)
5466 (vectorizable_reduction, vectorizable_induction): Pass back the
5467 vectorized statement as a stmt_vec_info.
5468 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5469 use of STMT_VINFO_VEC_STMT.
5470 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5471 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5472 as stmt_vec_infos rather than gimple stmts.
5473 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5475 (vectorizable_live_operation): Likewise.
5476 (vectorizable_reduction, vectorizable_induction): Likewise,
5477 updating use of STMT_VINFO_VEC_STMT.
5478 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5479 of STMT_VINFO_VEC_STMT.
5480 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5481 (vectorizable_simd_clone_call, vectorizable_conversion)
5482 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5483 (vectorizable_store, vectorizable_load, vectorizable_condition)
5484 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5485 from a gimple stmt to a stmt_vec_info.
5486 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5487 pointer to a stmt_vec_info to the vectorizable_* routines.
5489 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5491 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5492 a gimple stmt to a stmt_vec_info.
5493 (is_pattern_stmt_p): Update accordingly.
5494 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5495 (vect_record_grouped_load_vectors): Likewise.
5496 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5497 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5498 (vect_model_reduction_cost): Likewise.
5499 (vect_create_epilog_for_reduction): Likewise.
5500 (vectorizable_reduction, vectorizable_induction): Likewise.
5501 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5502 Return the stmt_vec_info for the pattern statement.
5503 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5504 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5505 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5506 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5507 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5508 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5509 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5510 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5514 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5515 (vect_finish_stmt_generation): Likewise.
5516 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5517 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5518 (vect_build_gather_load_calls): Use the return value of the above
5519 functions instead of a separate call to vinfo_for_stmt. Use narrow
5520 scopes for the input gimple stmt and wider scopes for the associated
5521 stmt_vec_info. Use vec_info::lookup_def when setting these
5522 stmt_vec_infos from an SSA_NAME definition.
5523 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5524 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5525 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5526 (vectorizable_store, vectorizable_load, vectorizable_condition)
5527 (vectorizable_comparison): Likewise.
5528 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5529 (vectorizable_reduction): Likewise.
5531 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5533 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5534 stmt_vec_info * parameter before the optional gimple **.
5535 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5536 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5537 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5538 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5539 (vectorizable_live_operation): Likewise.
5540 * tree-vect-patterns.c (type_conversion_p): Likewise.
5541 (vect_look_through_possible_promotion): Likewise.
5542 (vect_recog_rotate_pattern): Likewise.
5543 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5545 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5547 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5548 a typedef to a wrapper class.
5549 (NULL_STMT_VEC_INFO): New macro.
5550 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5551 (stmt_vec_info::operator*): New function.
5552 (stmt_vec_info::operator gimple *): Likewise.
5553 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5554 (add_stmt_costs): Likewise.
5555 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5556 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5557 (vect_get_known_peeling_cost): Likewise.
5558 (vect_estimate_min_profitable_iters): Likewise.
5559 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5560 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5561 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5562 (vectorizable_store, free_stmt_vec_infos): Likewise.
5563 (new_stmt_vec_info): Change return type of xcalloc to
5566 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5568 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5569 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5570 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5571 a single_imm_use-based sequence.
5572 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5574 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5576 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5577 * tree-vectorizer.c (vec_info::lookup_def): New function.
5578 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5579 (vect_widened_op_tree): Likewise.
5580 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5581 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5582 (vectorizable_reduction): Likewise.
5583 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5585 (vect_is_slp_reduction): Update calls accordingly. Use
5586 vec_info::lookup_def.
5587 (vect_is_simple_reduction): Likewise
5588 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5590 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5592 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5593 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5594 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5596 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5597 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5598 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5599 (vect_is_slp_reduction, vectorizable_induction)
5600 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5601 * tree-vect-patterns.c (vect_init_pattern_stmt):
5602 (vect_determine_min_output_precision_1, vect_determine_precisions)
5603 (vect_pattern_recog): Likewise.
5604 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5605 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5606 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5607 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5608 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5609 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5610 info field from a loop to a loop_vec_info.
5612 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5614 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5615 (vec_info::add_stmt): Declare.
5616 * tree-vectorizer.c (vec_info::add_stmt): New function.
5617 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5618 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5619 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5620 (vectorizable_induction): Likewise.
5621 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5622 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5623 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5624 (vectorizable_load): Likewise.
5625 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5626 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5627 (vect_recog_gather_scatter_pattern): Likewise.
5628 (append_pattern_def_seq): Likewise. Remove a check that is
5629 performed by add_stmt itself.
5631 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5633 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5634 which make_ssa_name was called with new_stmt before new_stmt
5637 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5639 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5641 (vect_is_slp_reduction): ...here...
5642 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5643 that are already known to be false.
5645 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5647 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5648 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5649 STMT_VINFO_NUM_SLP_USES when it's true.
5650 (vect_free_slp_instance): Add a final_p parameter and pass it to
5652 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5653 (vect_analyze_slp_instance): Likewise.
5654 (vect_slp_analyze_operations): Likewise.
5655 (vect_slp_analyze_bb_1): Likewise.
5656 * tree-vectorizer.c (vec_info): Likewise.
5657 * tree-vect-loop.c (vect_transform_loop): Likewise.
5659 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5661 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5662 function is not called for second and subsequent members of
5665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5667 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5668 cases for nested loops from here to ...
5669 (vect_create_epilog_for_reduction): ...here. Only call
5670 vect_is_simple_use for inner-loop reductions.
5672 2018-07-31 Martin Liska <mliska@suse.cz>
5674 PR gcov-profile/85338
5675 PR gcov-profile/85350
5676 PR gcov-profile/85372
5677 * profile.c (struct location_triplet): New.
5678 (struct location_triplet_hash): Likewise.
5679 (output_location): Do not output a BB that
5680 is already recorded for a line.
5681 (branch_prob): Use streamed_locations.
5683 2018-07-31 Martin Liska <mliska@suse.cz>
5685 PR gcov-profile/85370
5686 * coverage.c (coverage_begin_function): Do not mark target
5687 clones as artificial functions.
5689 2018-07-31 Martin Liska <mliska@suse.cz>
5691 PR gcov-profile/83813
5692 PR gcov-profile/84758
5693 PR gcov-profile/85217
5694 PR gcov-profile/85332
5695 * profile.c (branch_prob): Do not record GOTO expressions
5696 for GIMPLE statements which locations are already streamed.
5698 2018-07-31 Olivier Hainque <hainque@adacore.com>
5700 * gcc.c (handle_spec_function): Accept a soft_matched_part
5701 argument, as do_spec_1. Pass it down to ...
5702 (eval_spec_function): Accept a soft_matched_part argument,
5703 and pass it down to ...
5704 (do_spec_2): Accept a soft_matched_part argument, and pass
5705 it down to do_spec_1.
5706 (do_spec_1): Pass soft_matched_part to handle_spec_function.
5707 (handle_braces): Update call to handle_spec_function.
5708 (driver::set_up_specs): Update calls to do_spec_2.
5709 (compare_debug_dump_opt_spec_function): Likewise.
5710 (compare_debug_self_opt_spec_function): Likewise.
5712 2018-07-31 Olivier Hainque <hainque@adacore.com>
5714 * common.opt (nolibc): New option.
5715 * doc/invoke.texi (Link Options): Document it.
5716 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5717 * config/alpha/linux.h: Likewise.
5718 * config/arc/elf.h: Likewise.
5719 * config/arm/uclinux-elf.h: Likewise.
5720 * config/arm/unknown-elf.h: Likewise.
5721 * config/avr/avrlibc.h: Likewise.
5722 * config/bfin/bfin.h: Likewise.
5723 * config/bfin/linux.h: Likewise.
5724 * config/bfin/uclinux.h: Likewise.
5725 * config/darwin.h: Likewise.
5726 * config/darwin10.h: Likewise.
5727 * config/darwin12.h: Likewise.
5728 * config/gnu-user.h: Likewise.
5729 * config/lm32/uclinux-elf.h: Likewise.
5730 * config/pa/pa-hpux11.h: Likewise.
5731 * config/pa/pa64-hpux.h: Likewise.
5732 * config/sparc/sparc.h: Likewise.
5734 2018-07-31 Olivier Hainque <hainque@adacore.com>
5736 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5737 undefined variables.
5739 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5742 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5745 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5747 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5748 terminated string literal.
5750 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5752 PR rtl-optimization/85160
5753 * combine.c (is_just_move): New function.
5754 (try_combine): Allow combining two instructions into two if neither of
5755 the original instructions was a move.
5757 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
5760 * doc/extend.texi (Global Register Variables): Discourage use of type
5762 (Local Register Variables): Likewise.
5764 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
5766 PR tree-optimization/86506
5767 * hwint.h (ceil_log2): Resync with hwint.c implementation.
5769 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
5772 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5773 hard_regno, make sure no insn between `from` and `to` clobbers it.
5775 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
5776 Tom de Vries <tdevries@suse.de>
5778 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5779 (PTX_DEFAULT_RUNTIME_DIM): ... this.
5780 (nvptx_goacc_validate_dims): Set default worker and gang dims to
5781 PTX_DEFAULT_RUNTIME_DIM.
5782 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5784 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
5786 * config/pa/pa.c (pa_output_addr_vec): Align address table.
5787 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5788 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5790 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
5792 * config/rs6000/constraints.md (wG constraint): Delete, no longer
5794 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5795 predicate to reflect toc fusion has been deleted.
5796 (toc_fusion_mem_raw): Delete, no longer used.
5797 (toc_fusion_mem_wrapped): Likewise.
5798 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5800 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5801 Delete, no longer used.
5802 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5803 meant to be used for toc fusion.
5804 (rs6000_debug_print_mode): Delete toc fusion debugging.
5805 (rs6000_debug_reg_global): Likewise.
5806 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5807 fusion and secondary reload support that were never used.
5808 (rs6000_option_override_internal): Delete TOC fusion, that was only
5809 partially defined, and it did not work unless you also used the
5811 (rs6000_legitimate_address_p): Delete TOC fusion support.
5812 (rs6000_opt_masks): Likewise.
5813 (fusion_wrap_memory_address): Delete function, no longer used.
5814 (fusion_split_address); Delete TOC fusion support.
5815 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5816 longer used with toc fusion being deleted.
5817 (TARGET_TOC_FUSION_FP): Likewise.
5818 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5820 (toc fusion spliter): Delete TOC fusion support.
5821 (toc_fusionload_<mode>): Likewise.
5822 (toc_fusionload_di): Likewise.
5823 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5824 longer needs to be named. Rename predicate to delete TOC fusion.
5825 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5826 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5827 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5828 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5829 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5831 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5833 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5834 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5835 __int128 in built-in function prototypes.
5836 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5837 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5839 2018-07-27 Martin Sebor <msebor@redhat.com>
5841 PR tree-optimization/86696
5842 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5843 types, including enums.
5844 (handle_char_store): Be prepared for the above function to fail.
5846 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5848 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5849 when optimization level is lower than 2 or optimize for size.
5851 2018-07-26 Martin Sebor <msebor@redhat.com>
5853 PR tree-optimization/86043
5854 PR tree-optimization/86042
5855 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5856 (get_string_cst_length): Rename...
5857 (get_min_string_length): ...to this. Add argument.
5858 (handle_char_store): Extend to handle multi-character stores by
5860 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5861 * tree.h (initializer_zerop): Add argument.
5863 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5866 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5867 declare target to variables if they have always,{to,from,tofrom} map
5870 2018-07-26 Martin Liska <mliska@suse.cz>
5873 * lto-wrapper.c: Add linker_output as prefix
5874 for ltrans_output_file.
5876 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5878 PR rtl-optimization/85805
5879 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5880 value for hard registers if that was written in the same mode.
5882 2018-07-26 Martin Liska <mliska@suse.cz>
5884 PR gcov-profile/86536
5885 * gcov.c (format_gcov): Use printf format %.*f directly
5886 and do not handle special values.
5888 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5890 * common/config/arc/arc-common.c (arc_option_optimization_table):
5891 Update default optimizations for size.
5893 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5895 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5896 * config/arc/constraints.md (CfZ): New constraint.
5897 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5898 (subsf3_fpu): Likewise.
5899 (cmpsf_fpu): Likewise.
5900 (cmpsf_fpu_uneq): Likewise.
5902 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5904 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5906 (arc_is_uncached_mem_p): uncached applies to both the variable and
5909 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5911 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5914 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5916 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5917 field "m_scopes" from vec to auto_vec.
5919 2018-07-25 Martin Liska <mliska@suse.cz>
5921 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5924 2018-07-25 Richard Biener <rguenther@suse.de>
5927 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5928 special wrt context_die late.
5929 (gen_subprogram_die): Re-use DIEs in local scope.
5931 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5933 PR tree-optimization/86644
5934 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5936 2018-07-25 Martin Liska <mliska@suse.cz>
5939 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5940 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5942 2018-07-25 Martin Liska <mliska@suse.cz>
5945 * params.def: Explain ASan abbreviation and provide
5946 a documentation link.
5948 2018-07-24 Martin Sebor <msebor@redhat.com>
5950 PR tree-optimization/86622
5951 PR tree-optimization/86532
5952 * builtins.h (string_length): Declare.
5953 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5954 (check_access): Be prepared for non-constant length ranges.
5955 (string_length): Make extern.
5956 * expr.c (string_constant): Only handle the minor non-constant
5957 array index. Use string_constant to compute the length of
5958 a generic string constant.
5960 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5962 PR tree-optimization/86618
5963 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5964 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5966 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5968 PR tree-optimization/86636
5969 * json.cc (json::object::set): Fix comment. Add assertions.
5970 (json::array::append): Move here from json.h. Add comment and an
5972 (json::string::string): Likewise.
5973 * json.h (json::array::append): Move to json.cc.
5974 (json::string::string): Likewise.
5975 * optinfo-emit-json.cc
5976 (optrecord_json_writer::impl_location_to_json): Assert that we
5977 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5978 wrapper around it. Expand the location once, rather than three
5980 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5981 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5983 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5984 places. Fix some overlong lines.
5986 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5988 * config/aarch64/aarch64-simd.md
5989 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5990 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5991 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5992 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5993 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5994 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5995 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5996 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5997 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5999 2018-07-24 Jakub Jelinek <jakub@redhat.com>
6002 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
6003 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
6004 and abs_d == d, do the power of two handling if profitable.
6006 2018-07-24 Richard Biener <rguenther@suse.de>
6008 * match.pd: Add BIT_FIELD_REF canonicalizations.
6010 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6013 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
6015 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6017 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
6018 terminated STRING_CST object.
6020 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6022 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
6024 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6026 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6027 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
6028 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
6029 the elements into a register.
6030 (rs6000_split_v4si_init_di_reg): Delete.
6031 (rs6000_split_v4si_init): Delete.
6032 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
6033 (vsx_init_v4si): Rewrite as a define_expand.
6035 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6037 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
6038 zero_extend argument from memory): New.
6040 2018-07-22 Martin Sebor <msebor@redhat.com>
6043 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
6044 diagnosing calls with unknown arguments unless -Walloca-larger-than
6045 is restricted to less than PTRDIFF_MAX bytes.
6047 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
6049 * doc/gcov.texi (Invoking Gcov): Editorial changes.
6051 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6053 * pretty-print.c (text_info::set_location): Remove redundant
6054 "line_table" parameter from call to rich_location::set_range.
6056 2018-07-20 Martin Sebor <msebor@redhat.com>
6059 * builtins.c (expand_builtin_alloca): Adjust.
6060 * calls.c (alloc_max_size): Simplify.
6061 * cgraphunit.c (cgraph_node::expand): Adjust.
6062 * common.opt (larger_than_size, warn_frame_larger_than): Remove
6064 (frame_larger_than_size): Same.
6065 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
6066 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
6068 * doc/invoke.texi (GCC Command Options): Document option arguments.
6069 Explain byte-size arguments and suffixes.
6070 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
6071 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
6072 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
6073 * doc/options.texi (UInteger): Expand.
6074 (Host_Wide_Int, ByteSize): Document new properties.
6075 * final.c (final_start_function_1): Include sizes in an error message.
6076 * function.c (frame_offset_overflow): Same.
6077 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
6078 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
6079 Diagnose unbounded alloca calls only for limits of less than
6081 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
6082 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
6084 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
6085 only for limits of less than PTRDIFF_MAX.
6086 * langhooks-def.h (lhd_handle_option): Change function argument
6088 * langhooks.c (lhd_handle_option): Same.
6089 * langhooks.h (handle_option): Same.
6090 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
6092 (var_type, var_type_struct): Same.
6093 (var_set): Handle ByteSize flag.
6094 * optc-gen.awk: Add comments to output to ease debugging. Make
6095 use of HOST_WIDE_INT where appropriate.
6096 * opts-gen-save.awk: Use %lx to format unsigned long.
6097 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
6098 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
6099 arguments. Parse bytes-size suffixes.
6100 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
6101 (enum_value_to_arg): Same.
6102 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
6103 (handle_option): Adjust.
6104 (generate_option): Change function argument to HOST_WIDE_INT.
6105 (cmdline_handle_error): Adjust.
6106 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
6107 (set_option): Change function argument to HOST_WIDE_INT.
6108 (option_enabled): Handle cl_host_wide_int.
6109 (get_option_state): Handle CLVC_SIZE.
6110 (control_warning_option): Same.
6111 * opts.c (common_handle_option): Change function argument to
6112 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
6113 OPT_Wvla_larger_than_.
6114 * opts.h (enum cl_var_type): Add an enumerator.
6115 * stor-layout.c (layout_decl): Print a more meaningful warning.
6116 * toplev.c (output_stack_usage): Adjust.
6118 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
6120 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
6121 call to inline_expand_builtin_string_cmp.
6122 (expand_builtin_strcmp): Likewise.
6123 (expand_builtin_strncmp): Likewise.
6124 (inline_string_cmp): Delete the last parameter, change char_type_node
6125 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
6127 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
6128 the inlining expansion on target where the type of the call has same or
6129 narrower precision than unsigned char.
6131 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6133 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
6134 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
6135 * common.opt (fsave-optimization-record): New option.
6136 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
6137 * doc/invoke.texi (-fsave-optimization-record): New option.
6138 * dumpfile.c: Include "optinfo-emit-json.h".
6139 (struct kv_pair): Move to coretypes.h.
6140 (optgroup_options): Make non-static.
6141 (dump_context::end_scope): Call
6142 optimization_records_maybe_pop_dump_scope.
6143 * dumpfile.h (optgroup_options): New decl.
6144 * json.cc: New file.
6146 * optinfo-emit-json.cc: New file.
6147 * optinfo-emit-json.h: New file.
6148 * optinfo.cc: Include "optinfo-emit-json.h".
6149 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
6150 (optinfo_enabled_p): Check optimization_records_enabled_p.
6151 (optinfo_wants_inlining_info_p): Likewise.
6152 * optinfo.h: Update comment.
6153 * profile-count.c (profile_quality_as_string): New function.
6154 * profile-count.h (profile_quality_as_string): New decl.
6155 (profile_count::quality): New accessor.
6156 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
6157 and optinfo_emit_json_cc_tests.
6158 * selftest.h (selftest::json_cc_tests): New decl.
6159 (selftest::optinfo_emit_json_cc_tests): New decl.
6160 * toplev.c: Include "optinfo-emit-json.h".
6161 (compile_file): Call optimization_records_finish.
6162 (do_compile): Call optimization_records_start.
6163 * tree-ssa-live.c: Include optinfo.h.
6164 (remove_unused_scope_block_p): Retain inlining information if
6165 optinfo_wants_inlining_info_p returns true.
6167 2018-07-20 Richard Biener <rguenther@suse.de>
6170 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
6171 to cover -flto-partition=none.
6173 2018-07-20 Martin Liska <mliska@suse.cz>
6175 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
6176 (get_decl_source_range): Remove unused function.
6178 2018-07-20 Richard Biener <rguenther@suse.de>
6180 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
6181 (struct vn_phi_s): Likewise.
6182 (struct vn_reference_s): Likewise.
6183 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
6184 for searching the slot of an entry known to be in the hash itself.
6185 (vn_phi_hasher::equal): Likewise.
6186 (vn_reference_hasher::equal): Likewise.
6187 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
6189 (optimistic_info, current_info): Remove, keeping only valid_info.
6190 (vn_reference_lookup_1): Remove fallback lookup.
6191 (vn_reference_lookup_2): Likewise.
6192 (vn_nary_op_lookup_1): Likewise.
6193 (vn_phi_lookup): Likewise.
6194 (vn_nary_build_or_lookup_1): Make sure to not chain the built
6196 (vn_reference_insert): Adjust, chain the inserted hash element
6197 at last_inserted_ref.
6198 (vn_reference_insert_pieces): Likewise.
6199 (visit_reference_op_call): Likewise.
6200 (vn_nary_op_insert_into): Chain the inserted hash element at
6202 (vn_nary_op_insert_pieces): Adjust.
6203 (vn_nary_op_insert): Likewise.
6204 (vn_nary_op_insert_stmt): Likewise.
6205 (vn_phi_insert): Adjust, chain the inserted hash element at
6207 (process_scc): Remove clearing and copying the optimistic
6208 table. Instead remove elements inserted during an optimistic
6209 iteration from the single table we maintain.
6210 (init_scc_vn): Adjust.
6211 (free_scc_vn): Likewise.
6212 (sccvn_dom_walker::record_cond): Likewise.
6213 (sccvn_dom_walker::after_dom_children): Likewise.
6215 2018-07-19 Martin Sebor <msebor@redhat.com>
6217 PR tree-optimization/84047
6218 PR tree-optimization/83776
6219 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6220 (check_array_bounds): Call it.
6222 2018-07-19 Martin Sebor <msebor@redhat.com>
6224 * align.h (align_flags): Use member initialization.
6226 2018-07-19 David Malcolm <dmalcolm@redhat.com>
6228 * Makefile.in (OBJS): Add optinfo.o.
6229 * coretypes.h (class symtab_node): New forward decl.
6230 (struct cgraph_node): New forward decl.
6231 (class varpool_node): New forward decl.
6232 * dump-context.h: New file.
6233 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
6235 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
6236 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
6237 (set_alt_dump_file): Likewise.
6238 (dump_context::~dump_context): New dtor.
6239 (dump_gimple_stmt): Move implementation to...
6240 (dump_context::dump_gimple_stmt): ...this new member function.
6241 Add the stmt to any pending optinfo, creating one if need be.
6242 (dump_gimple_stmt_loc): Move implementation to...
6243 (dump_context::dump_gimple_stmt_loc): ...this new member function.
6244 Start a new optinfo and add the stmt to it.
6245 (dump_gimple_expr): Move implementation to...
6246 (dump_context::dump_gimple_expr): ...this new member function.
6247 Add the stmt to any pending optinfo, creating one if need be.
6248 (dump_gimple_expr_loc): Move implementation to...
6249 (dump_context::dump_gimple_expr_loc): ...this new member function.
6250 Start a new optinfo and add the stmt to it.
6251 (dump_generic_expr): Move implementation to...
6252 (dump_context::dump_generic_expr): ...this new member function.
6253 Add the tree to any pending optinfo, creating one if need be.
6254 (dump_generic_expr_loc): Move implementation to...
6255 (dump_context::dump_generic_expr_loc): ...this new member
6256 function. Add the tree to any pending optinfo, creating one if
6258 (dump_printf): Move implementation to...
6259 (dump_context::dump_printf_va): ...this new member function. Add
6260 the text to any pending optinfo, creating one if need be.
6261 (dump_printf_loc): Move implementation to...
6262 (dump_context::dump_printf_loc_va): ...this new member function.
6263 Start a new optinfo and add the stmt to it.
6264 (dump_dec): Move implementation to...
6265 (dump_context::dump_dec): ...this new member function. Add the
6266 value to any pending optinfo, creating one if need be.
6267 (dump_context::dump_symtab_node): New member function.
6268 (dump_context::get_scope_depth): New member function.
6269 (dump_context::begin_scope): New member function.
6270 (dump_context::end_scope): New member function.
6271 (dump_context::ensure_pending_optinfo): New member function.
6272 (dump_context::begin_next_optinfo): New member function.
6273 (dump_context::end_any_optinfo): New member function.
6274 (dump_context::s_current): New global.
6275 (dump_context::s_default): New global.
6276 (dump_scope_depth): Delete global.
6277 (dumpfile_ensure_any_optinfo_are_flushed): New function.
6278 (dump_symtab_node): New function.
6279 (get_dump_scope_depth): Reimplement in terms of dump_context.
6280 (dump_begin_scope): Likewise.
6281 (dump_end_scope): Likewise.
6282 (selftest::temp_dump_context::temp_dump_context): New ctor.
6283 (selftest::temp_dump_context::~temp_dump_context): New dtor.
6284 (selftest::verify_item): New function.
6285 (ASSERT_IS_TEXT): New macro.
6286 (ASSERT_IS_TREE): New macro.
6287 (ASSERT_IS_GIMPLE): New macro.
6288 (selftest::test_capture_of_dump_calls): New test.
6289 (selftest::dumpfile_c_tests): Call it.
6290 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
6291 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
6292 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
6293 descriptive comment.
6294 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
6295 (dump_node, dump_bb): Move these unrelated decls.
6296 (class dump_manager): Add leading comment.
6297 * optinfo.cc: New file.
6298 * optinfo.h: New file.
6300 2018-07-19 Michael Collison <michael.collison@arm.com>
6301 Richard Henderson <rth@redhat.com>
6303 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
6304 (subti): Handle op1 zero.
6305 (subvti4, usub4ti4): New.
6306 (*sub<GPI>3_compare1_imm): New.
6307 (sub<GPI>3_carryinCV): New.
6308 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
6309 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
6311 2018-07-19 Michael Collison <michael.collison@arm.com>
6312 Richard Henderson <rth@redhat.com>
6314 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
6315 (addti3): Create simpler code if low part is already known to be 0.
6316 (addvti4, uaddvti4): New.
6317 (*add<GPI>3_compareC_cconly_imm): New.
6318 (*add<GPI>3_compareC_cconly): New.
6319 (*add<GPI>3_compareC_imm): New.
6320 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
6321 handle constants within this pattern..
6322 (*add<GPI>3_compareV_cconly_imm): New.
6323 (*add<GPI>3_compareV_cconly): New.
6324 (*add<GPI>3_compareV_imm): New.
6325 (add<GPI>3_compareV): New.
6326 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
6327 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
6328 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
6329 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
6330 with 'comparison' operator.
6331 (*add<GPI>3_compareV_cconly_imm): Ditto.
6332 (*add<GPI>3_compareV_cconly): Ditto.
6333 (*add<GPI>3_compareV_imm): Ditto.
6334 (add<GPI>3_compareV): Ditto.
6335 (add<mode>3_carryinC): Ditto.
6336 (*add<mode>3_carryinC_zero): Ditto.
6337 (*add<mode>3_carryinC): Ditto.
6338 (add<mode>3_carryinV): Ditto.
6339 (*add<mode>3_carryinV_zero): Ditto.
6340 (*add<mode>3_carryinV): Ditto.
6342 2018-07-19 Michael Collison <michael.collison@arm.com>
6343 Richard Henderson <rth@redhat.com>
6345 * config/aarch64/aarch64-modes.def (CC_V): New.
6346 * config/aarch64/aarch64-protos.h
6347 (aarch64_addti_scratch_regs): Declare
6348 (aarch64_subvti_scratch_regs): Declare.
6349 (aarch64_expand_subvti): Declare.
6350 (aarch64_gen_unlikely_cbranch): Declare
6351 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
6352 for signed overflow using CC_Vmode.
6353 (aarch64_get_condition_code_1): Handle CC_Vmode.
6354 (aarch64_gen_unlikely_cbranch): New function.
6355 (aarch64_addti_scratch_regs): New function.
6356 (aarch64_subvti_scratch_regs): New function.
6357 (aarch64_expand_subvti): New function.
6359 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6361 * config/aarch64/aarch64-option-extensions.def: New entry for profile
6363 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
6364 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
6367 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6370 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6371 address check not strict.
6373 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6375 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
6377 (store_pair_lanes<mode>): Likewise.
6378 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
6379 enum value 'ADDR_QUERY_LDP_STP_N'.
6380 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
6381 (aarch64_print_address_internal): Add declaration.
6382 (aarch64_print_ldpstp_address): Remove.
6383 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
6384 (aarch64_print_operand): Change printing of 'y'.
6385 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
6386 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
6387 'true' rather than '1'.
6388 * config/aarch64/constraints.md (Uml): Likewise.
6389 (Uml): Rename to Umn.
6392 2018-07-19 Richard Biener <rguenther@suse.de>
6394 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
6396 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
6397 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
6398 (vn_reference_hasher): Likewise.
6399 (struct vn_tables_s): Remove obstack and alloc-pool members.
6400 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
6401 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
6402 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
6403 (vn_reference_insert_pieces): Likewise.
6404 (alloc_vn_nary_op_noinit): Adjust.
6405 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
6406 (vn_phi_eq): Adjust.
6407 (shared_lookup_phiargs): Remove.
6408 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
6409 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
6410 (visit_reference_op_call): Likewise.
6411 (copy_nary, copy_phi, copy_reference): Remove.
6412 (process_scc): Rewind the obstack when iterating. Do not
6413 copy the elements to valid_info but just move them from one
6414 hashtable to the other.
6415 (allocate_vn_table): Adjust.
6416 (free_vn_table): Likewise.
6417 (init_scc_vn): Likewise.
6418 (free_scc_vn): Likewise.
6420 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
6423 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
6424 indirect_return as function type attribute.
6425 (ix86_attribute_table): Change indirect_return to function
6427 * doc/extend.texi: Update indirect_return attribute.
6429 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
6431 * wide-int.h (widest2_int): New.
6432 * gimple-fold.c (arith_overflowed_p): Use it.
6433 * tree.h (widest2_int_cst): New.
6434 * tree-vrp.c (wide_int_binop_overflow): Rename from
6435 vrp_int_const_binop.
6436 Rewrite to work on trees.
6437 (extract_range_from_multiplicative_op_1): Abstract code to...
6438 (wide_int_range_min_max): ...here.
6439 (wide_int_range_cross_product): ...and here.
6440 (extract_range_from_binary_expr_1): Abstract overflow code to...
6441 (wide_int_range_mult_wrapping): ...here.
6442 * tree-vrp.h (wide_int_range_cross_product): New.
6443 (wide_int_range_mult_wrapping): New.
6445 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
6446 Julia Koval <julia.koval@intel.com>
6448 * config/i386/x86-tune-costs.h (skylake_memcpy,
6449 skylake_memset): Replace rep_prefix with unrolling for size 512.
6451 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
6454 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6455 comparision with EQ_EXPR in last stmt.
6457 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
6459 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6460 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6461 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6462 previously known as "PowerPC AltiVec Built-in Functions". Move
6463 some material to new subsubsections "PowerPC AltiVec Built-in
6464 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6466 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6467 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6468 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6469 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6471 2018-07-18 Richard Biener <rguenther@suse.de>
6473 PR tree-optimization/86557
6474 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6477 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6479 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6481 2018-07-17 Jeff Law <law@redhat.com>
6483 * config/arm/arm.c (get_label_padding): Update for recent
6484 changes to label_to_alignment.
6486 PR tree-optimization/86010
6487 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6489 * config/mips/mips.c (vr4130_align_insns): Update for recent
6490 changes to label_to_alignment.
6492 * config/frv/frv.c (frv_label_align): Update for recent changes
6493 to label_to_alignment.
6495 * config/nios2/nios2.c (nios2_label_align): Update for recent
6496 changes which dropped ALIGN_LABELS_LOG.
6498 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6500 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6501 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6503 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6505 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6506 deprecated align_labels_log.
6508 2018-07-17 Richard Biener <rguenther@suse.de>
6511 * dwarf2out.c (init_sections_and_labels): Always generate
6512 a debug_line_str_section for early LTO debug.
6513 (dwarf2out_finish): Reset debug_line_str_hash output early.
6514 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6516 (dwarf2out_early_finish): Output debug_line_str.
6518 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6520 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6521 index register on z196 or later.
6523 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6525 * config/s390/s390.c (s390_default_align): Set default function
6527 (s390_override_options_after_change): Call s390_default align.
6528 (s390_option_override_internal): Call s390_default_align.
6529 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6531 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6534 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6535 to _looptemp_ clauses, other than the first two.
6537 2018-07-17 Martin Liska <mliska@suse.cz>
6539 * opts.c: Do not enable OPT_falign_* for -Os.
6541 2018-07-17 Martin Liska <mliska@suse.cz>
6543 * align.h (MAX_CODE_ALIGN): New.
6544 (MAX_CODE_ALIGN_VALUE): New.
6545 * common/config/i386/i386-common.c (ix86_handle_option):
6546 (MAX_CODE_ALIGN): Moved to align.h.
6547 * final.c (MAX_CODE_ALIGN): Likewise.
6548 * opts.c (parse_and_check_align_values):
6549 (MAX_CODE_ALIGN): Likewise.
6550 (MAX_CODE_ALIGN_VALUE): Likewise.
6552 2018-07-17 Martin Liska <mliska@suse.cz>
6554 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6555 in order to fulfil coding style.
6556 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6557 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6558 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6559 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6560 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6561 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6562 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6563 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6564 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6566 2018-07-17 Martin Liska <mliska@suse.cz>
6568 * align.h: New file.
6569 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6571 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6572 align_flags of label_to_alignment.
6573 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6575 * config/m68k/m68k.c: Do not use removed align_labels_value and
6577 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6578 (LOOP_ALIGN): Likewise.
6579 (LABEL_ALIGN): Likewise.
6580 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6581 Remove not used macro.
6582 (rs6000_loop_align): Change return type to align_flags.
6583 (rs6000_loop_align_max_skip): Remove.
6584 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6585 Change return type to align_flags.
6586 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6587 Remove not used macro.
6588 (rs6000_loop_align): Change return type to align_flags.
6589 (rs6000_loop_align_max_skip): Remove.
6590 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6591 * config/rx/rx-protos.h (rx_align_for_label): Make it
6593 * config/rx/rx.c (rx_align_for_label): Change return type
6595 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6597 into align_flags class.
6598 (LABEL_ALIGN): Likewise.
6599 (LOOP_ALIGN): Likewise.
6600 * config/s390/s390.c (s390_label_align): Use align_flags
6602 (s390_asm_output_function_label): Likewise.
6603 * config/sh/sh.c (sh_override_options_after_change):
6604 Use align_flags class directly without macros.
6605 (find_barrier): Likewise.
6606 (barrier_align): Likewise.
6607 (sh_loop_align): Likewise.
6608 * config/spu/spu.c (spu_option_override):
6609 Use align_flags_tuple::get_value instead of removed macros.
6610 (spu_sched_init): Likewise.
6611 * config/spu/spu.h (GTY): Likewise.
6612 * config/visium/visium.c (visium_option_override):
6613 Set "8" as default secondary alignment.
6614 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6615 in order to guarantee secondary alignment of 8.
6616 * coretypes.h: Include align.h header file.
6617 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6618 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6619 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6620 * doc/tm.texi.in: Likewise.
6621 * final.c (struct label_alignment): Remove not used structure.
6622 (LABEL_ALIGN): Change type to align_flags.
6623 (LOOP_ALIGN): Likewise.
6624 (JUMP_ALIGN): Likewise.
6625 (default_loop_align_max_skip): Remove.
6626 (default_label_align_max_skip): Likewise.
6627 (default_jump_align_max_skip): Likewise.
6628 (default_label_align_after_barrier_max_skip):
6629 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6630 (LABEL_TO_MAX_SKIP): Remove.
6631 (label_to_alignment): Return align_flags type instead of integer.
6632 (label_to_max_skip): Remove.
6633 (align_fuzz): Use align_flags type.
6634 (compute_alignments): Use align_flags type and use align_flags::max
6635 to combine multiple alignments.
6636 (grow_label_align): Grow vec instead of C array.
6637 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6638 (shorten_branches): Use align_flags type and use align_flags::max
6639 to combine multiple alignments.
6640 (final_scan_insn_1): Remove usage of secondary alignment that comes
6641 from label alignment, but instead use proper secondary alignment
6642 which is computed in grow_label_align.
6643 * flags.h (struct align_flags_tuple): Move to align.h.
6644 (struct align_flags): Likewise.
6645 (state_align_loops): Rename to align_loops.
6646 (state_align_jumps): Rename to align_jumps.
6647 (state_align_labels): Rename to align_labels.
6648 (state_align_functions): Rename to align_functions.
6649 (align_loops_log): Remove.
6650 (align_jumps_log): Remove.
6651 (align_labels_log): Remove.
6652 (align_functions_log): Remove.
6653 (align_loops_max_skip): Remove.
6654 (align_jumps_max_skip): Remove.
6655 (align_labels_max_skip): Remove.
6656 (align_functions_max_skip): Remove.
6657 (align_loops_value): Remove.
6658 (align_jumps_value): Remove.
6659 (align_labels_value): Remove.
6660 (align_functions_value): Remove.
6661 * output.h (label_to_alignment): Change return type to align_flags.
6662 (label_to_max_skip): Remove.
6663 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6664 jump_align_max_skip macros.
6665 * targhooks.h (default_loop_align_max_skip): Remove.
6666 (default_label_align_max_skip): Likewise.
6667 (default_jump_align_max_skip): Likewise.
6668 (default_label_align_after_barrier_max_skip): Remove.
6669 * toplev.c (read_log_maxskip): Use ::normalize function.
6670 (parse_N_M): Remove not used argument and also call ::normalize.
6671 (parse_alignment_opts): Do not pass unused arguments.
6672 * varasm.c (assemble_start_function): Use directly align_functions
6673 instead of removed macros.
6674 * system.h: Do not poison removed macros.
6676 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6679 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6680 and cond temporaries don't have reference type if iterator has
6681 pointer type. For init use &for_pre_body instead of pre_p if
6682 for_pre_body is non-empty.
6684 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6686 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6687 double-double modes to SFmode directly directly.
6688 (trunc<mode>sf2_fprs): Delete.
6690 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6692 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6693 for conversions between IFmode and the decimal floating point modes.
6694 (init_float128_ieee): Use the correct names for conversions between
6695 KFmode and the decimal floating point modes.
6697 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6699 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6700 for the conversions between TDmode and IFmode.
6701 (init_float128_ieee): Use more correct names for the conversions
6702 between TDmode and KFmode.
6704 2018-07-16 Jakub Jelinek <jakub@redhat.com>
6706 PR tree-optimization/86526
6707 * builtins.c (expand_builtin_memcmp): Formatting fixes.
6708 (inline_expand_builtin_string_cmp): Likewise.
6709 (inline_string_cmp): Likewise. Use c_readstr instead of
6710 builtin_memcpy_read_str. Add unit_mode temporary.
6712 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6715 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6716 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6718 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6720 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6721 Alphabetize prototypes of built-in functions, separating out
6722 built-in functions that are listed in this section but should be
6723 described elsewhere.
6725 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
6728 * expmed.c (emit_store_flag): Do not emit setcc followed by a
6729 conditional move when trapping comparison was split to a
6730 non-trapping one (and vice versa).
6732 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6734 * config/s390/s390.c (s390_function_profiler): Generate nops
6735 instead of profiler call sequences.
6736 * config/s390/s390.opt: Add the new option.
6738 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6740 * config/s390/s390.c (s390_function_profiler): Generate
6741 __mcount_loc section.
6742 * config/s390/s390.opt: Add the new option.
6744 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6746 * common.opt: Add the new warning.
6747 * config/s390/s390.c (s390_function_profiler): Emit "brasl
6748 %r0,__fentry__" when -mfentry is specified.
6749 (s390_option_override_internal): Disallow -mfentry for 31-bit
6751 * config/s390/s390.opt: Add the new option.
6753 2018-07-16 Richard Biener <rguenther@suse.de>
6756 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6757 for function-local FUNCTION_DECL and RESULT_DECL immediately.
6759 2018-07-16 Martin Liska <mliska@suse.cz>
6762 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6765 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6767 * config/arc/arcHS.md: Update ARCHS scheduling rules.
6769 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6771 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6773 * config/arc/arc-cpus.def (hs4x): New cpu.
6775 * config/arc/arc-tables.opt: Regenerate.
6776 * config/arc/arc.c (arc_sched_issue_rate): New function.
6777 (TARGET_SCHED_ISSUE_RATE): Define.
6778 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6779 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6781 (attr tune): Add ARCHS4x tune values.
6782 (attr tune_dspmpy): Define.
6783 (*tst): Correct instruction type.
6784 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6785 * config/arc/arcHS4x.md: New file.
6786 * config/arc/fpu.md: Update instruction type attributes.
6787 * config/arc/t-multilib: Regenerate.
6789 2018-07-16 Tom de Vries <tdevries@suse.de>
6792 * var-tracking.c (vt_initialize): Fix pre_dec handling.
6794 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6796 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6799 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
6801 PR tree-optimization/86514
6802 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6803 conversion to a boolean type from a type with greater precision.
6805 2018-07-16 Tom de Vries <tdevries@suse.de>
6807 * var-tracking.c (vt_initialize): Print adjusted insn slim if
6808 dump_flags request TDF_SLIM.
6810 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6812 * fold-const.c (int_const_binop_1): Abstract...
6813 (wide_int_binop): ...wide int code here.
6814 (poly_int_binop): ...poly int code here.
6815 Abstract the rest of int_const_binop_1 into int_const_binop.
6816 * fold-const.h (wide_int_binop): New.
6817 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6818 Remove useless PLUS/MINUS_EXPR case.
6819 (zero_nonzero_bits_from_vr): Move wide int code...
6820 (zero_nonzero_bits_from_bounds): ...here.
6821 (extract_range_from_binary_expr_1): Move mask optimization code...
6822 (range_easy_mask_min_max): ...here.
6823 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6824 (range_easy_mask_min_max): New.
6826 2018-07-15 Jeff Law <law@redhat.com>
6829 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6832 2018-07-14 Jim Wilson <jimw@sifive.com>
6834 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6836 2018-07-14 Paul Koning <ni1d@arrl.net>
6838 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6840 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6842 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6845 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6846 Steve Munroe <munroesj52@gmail.com>
6848 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6849 (_mm_andnot_si128): Likewise.
6850 (_mm_or_si128): Likewise.
6851 (_mm_xor_si128): Likewise.
6853 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6856 * builtins.c (expand_builtin_memcmp): Inline the calls first
6857 when result_eq is false.
6858 (expand_builtin_strcmp): Inline the calls first.
6859 (expand_builtin_strncmp): Likewise.
6860 (inline_string_cmp): New routine. Expand a string compare
6861 call by using a sequence of char comparison.
6862 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6863 a call to str(n)cmp/memcmp.
6864 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6866 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6868 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6870 * config/arm/driver-arm.c: Include arm-native.h.
6871 (host_detect_local_cpu): Use auto-generated data tables.
6872 (vendors, arm_cpu_table): Delete. Move part information to ...
6873 * config/arm/arm-cpus.in: ... here.
6874 * config/arm/parsecpu.awk (gen_native): New function.
6875 (vendor, part): New CPU fields.
6876 (END): Add support for building the native CPU detection tables.
6877 * config/arm/t-arm (arm-native.h): Add build rule.
6878 (driver-arm.o): Add dependency on arm-native.h.
6880 2018-07-13 Richard Biener <rguenther@suse.de>
6883 * match.pd (addr1 - addr2): Allow either of the operand to
6886 2018-07-13 Tom de Vries <tdevries@suse.de>
6888 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6891 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6893 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6894 arrays instead of numbered variables.
6896 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6898 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6899 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6901 2018-07-13 Richard Biener <rguenther@suse.de>
6904 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6905 instead of get_context_die.
6907 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6908 Richard Biener <rguenther@suse.de>
6911 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6912 that the loop latch destination where phi is defined.
6914 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6916 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6917 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6918 (riscv_set_current_function): Extract getting interrupt type to new
6920 (riscv_get_interrupt_type): New function.
6921 (riscv_merge_decl_attributes): New function, checking interrupt type is
6923 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6925 2018-07-12 Paul Koning <ni1d@arrl.net>
6927 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6930 2018-07-12 Paul Koning <ni1d@arrl.net>
6932 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6933 zero reference, add doloop_end instead.
6934 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6935 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6936 detail for doloop_end.
6938 2018-07-12 Martin Sebor <msebor@redhat.com>
6941 * attribs.c (decl_attributes): Reject conflicting attributes before
6942 calling attribute handlers.
6944 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6946 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6948 (gcc::dump_manager::get_dump_file_name): likewise.
6949 (dump_begin): Likewise.
6950 * dumpfile.h (dump_begin): Update prototype.
6951 (gcc::dump_manager::get_dump_file_name,
6952 gcc::dump_manager::get_dump_file_name): Update prototype.
6954 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6956 * internal-fn.h (vectorizable_internal_fn_p): New function.
6957 * tree-vect-slp.c (compatible_calls_p): Likewise.
6958 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6959 to internal functions.
6960 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6962 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6964 * fold-const.h (inverse_conditions_p): Declare.
6965 * fold-const.c (inverse_conditions_p): New function.
6966 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6967 that test the inverse condition of a conditional internal function.
6968 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6969 * internal-fn.c (internal_fn_mask_index): Handle conditional
6971 (vectorized_internal_fn_supported_p): New function.
6972 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6973 (any_pred_load_store): Replace with...
6974 (need_to_predicate): ...this new variable.
6975 (redundant_ssa_names): New variable.
6976 (ifcvt_can_use_mask_load_store): Move initial checks to...
6977 (ifcvt_can_predicate): ...this new function. Handle tree codes
6978 for which a conditional internal function exists.
6979 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6980 instead of ifcvt_can_use_mask_load_store. Update after variable
6982 (predicate_load_or_store): New function, split out from
6983 predicate_mem_writes.
6984 (check_redundant_cond_expr): New function.
6985 (value_available_p): Likewise.
6986 (predicate_rhs_code): Likewise.
6987 (predicate_mem_writes): Rename to...
6988 (predicate_statements): ...this. Use predicate_load_or_store
6989 and predicate_rhs_code.
6990 (combine_blocks, tree_if_conversion): Update after above name changes.
6991 (ifcvt_local_dce): Handle redundant_ssa_names.
6992 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6993 general conditional functions.
6994 * tree-vect-stmts.c (vectorizable_call): Likewise.
6996 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6997 Alan Hayward <alan.hayward@arm.com>
6998 David Sherwood <david.sherwood@arm.com>
7000 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
7001 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7002 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
7003 plus and minus and convert them into IFN_COND_FMA-based sequences.
7004 (convert_mult_to_fma): Handle conditional plus and minus.
7006 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7008 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
7009 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
7010 (cond_fnms_optab): New optabs.
7011 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
7013 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
7014 * internal-fn.h (get_conditional_internal_fn): Declare.
7015 (get_unconditional_internal_fn): Likewise.
7016 * internal-fn.c (cond_ternary_direct): New macro.
7017 (expand_cond_ternary_optab_fn): Likewise.
7018 (direct_cond_ternary_optab_supported_p): Likewise.
7019 (FOR_EACH_COND_FN_PAIR): Likewise.
7020 (get_conditional_internal_fn): New function.
7021 (get_unconditional_internal_fn): Likewise.
7022 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
7023 (gimple_match_op::gimple_match_op): Add a new overload for 5
7025 (gimple_match_op::set_op): Likewise.
7026 (gimple_resimplify5): Declare.
7027 * genmatch.c (decision_tree::gen): Generate simplifications for
7029 * gimple-match-head.c (gimple_simplify): Define an overload for
7030 5 operands. Handle calls with 5 arguments in the top-level overload.
7031 (convert_conditional_op): Handle conversions from unconditional
7032 internal functions to conditional ones.
7033 (gimple_resimplify5): New function.
7034 (build_call_internal): Pass a fifth operand.
7035 (maybe_push_res_to_seq): Likewise.
7036 (try_conditional_simplification): Try converting conditional
7037 internal functions to unconditional internal functions.
7038 Handle 3-operand unconditional forms.
7039 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
7040 Define ternary equivalents of the current rules for binary conditional
7042 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
7044 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
7045 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
7046 (optab): Handle them.
7047 (SVE_COND_FP_TERNARY): New int iterator.
7048 (sve_fmla_op, sve_fmad_op): New int attributes.
7049 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
7050 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
7051 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
7053 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7055 * target.def (preferred_else_value): New target hook.
7056 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
7057 * doc/tm.texi: Regenerate.
7058 * targhooks.h (default_preferred_else_value): Declare.
7059 * targhooks.c (default_preferred_else_value): New function.
7060 * internal-fn.h (conditional_internal_fn_code): Declare.
7061 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
7062 (get_conditional_internal_fn): Use it.
7063 (conditional_internal_fn_code): New function.
7064 * gimple-match.h (gimple_match_cond): New struct.
7065 (gimple_match_op): Add a cond member function.
7066 (gimple_match_op::gimple_match_op): Update all forms to take a
7068 * genmatch.c (expr::gen_transform): Use the same condition as res_op
7069 for the suboperation, but don't specify a particular else_value.
7070 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
7071 (visit_nary_op, visit_reference_op_load): Pass
7072 gimple_match_cond::UNCOND to the gimple_match_op constructor.
7073 * gimple-match-head.c: Include tree-eh.h
7074 (convert_conditional_op): New function.
7075 (maybe_resimplify_conditional_op): Likewise.
7076 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
7077 (gimple_resimplify2): Likewise.
7078 (gimple_resimplify3): Likewise.
7079 (gimple_resimplify4): Likewise.
7080 (maybe_push_res_to_seq): Return null for conditional operations.
7081 (try_conditional_simplification): New function.
7082 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
7084 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
7086 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
7088 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
7090 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7092 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7094 (hash_tree): Do not hash DECL_FCONTEXT
7095 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7096 Do not stream DECL_FCONTEXT.
7097 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
7098 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
7100 2018-07-12 Richard Biener <rguenther@suse.de>
7103 * dwarf2out.c (gen_block_die): Only output blocks when they have
7104 at least one !DECL_IGNORED_P variable.
7106 2018-07-12 Richard Biener <rguenther@suse.de>
7109 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
7110 Remove -mieee-fp handling.
7112 2018-07-12 Richard Biener <rguenther@suse.de>
7114 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
7115 left-over from last patch.
7117 2018-07-12 Jakub Jelinek <jakub@redhat.com>
7119 PR tree-optimization/86492
7120 * gimple-ssa-store-merging.c
7121 (imm_store_chain_info::coalesce_immediate_stores): Call
7122 check_no_overlap even for the merge_overlapping case. Formatting fix.
7124 2018-07-12 Richard Biener <rguenther@suse.de>
7127 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
7128 move possibly trapping operations into the conditional.
7130 2018-07-12 Richard Biener <rguenther@suse.de>
7132 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
7133 (vn_lookup_simplify_result): Remove recursion limit applied
7135 (vn_nary_build_or_lookup_1): Adjust.
7136 (try_to_simplify): Likewise.
7137 * gimple-match-head.c (gimple_resimplify1): Instead apply one
7139 (gimple_resimplify2): Likewise.
7140 (gimple_resimplify3): Likewise.
7141 (gimple_resimplify4): Likewise.
7143 2018-07-11 Jakub Jelinek <jakub@redhat.com>
7145 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
7146 Use __mmask64 type instead of __mmask8 for __M argument.
7147 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
7148 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
7150 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
7151 __mmask16 for __M argument.
7152 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
7153 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
7154 to __mmask16 instead of __mmask8.
7155 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
7156 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
7157 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
7158 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
7159 instead of __mmask16 for __U argument.
7160 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
7161 __mmask16 instead of __mmask8 for __U argument.
7162 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
7164 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
7166 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
7168 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
7170 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
7172 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
7174 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
7176 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
7178 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
7180 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
7182 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
7183 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
7184 return type as well as __M argument type and all casts from __mmask8
7186 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
7187 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
7188 return type as well as __M argument type and all casts from __mmask8
7190 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
7191 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
7192 return type as well as __M argument type and all casts from __mmask8
7194 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
7195 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
7196 return type as well as __M argument type and all casts from __mmask8
7198 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
7199 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
7202 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
7204 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
7205 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
7208 2018-07-11 Paul Koning <ni1d@arrl.net>
7210 * doc/md.texi (define_subst): Document how multiple occurrences of
7211 the same argument in the replacement pattern are handled.
7213 2018-07-11 Paul Koning <ni1d@arrl.net>
7215 * doc/extend.texi (Common Variable Attributes): Move "mode" into
7217 (Common Type Attributes): Add "mode" attribute.
7219 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
7221 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
7222 stream DECL_ORIGINAL_TYPE.
7223 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
7224 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
7225 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
7226 Do not walk original type.
7227 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
7229 (write_ts_decl_non_common_tree_pointers): Do not stream
7231 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
7232 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
7234 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7236 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
7237 thread twice from the same starting edge.
7239 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7241 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
7242 * gimple.c (gimple_call_nonnull_result_p): ...here...
7243 (gimple_call_nonnull_arg): ...and here.
7244 * gimple.h (gimple_call_nonnull_result_p): New.
7245 (gimple_call_nonnull_arg): New.
7247 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
7249 * config/arm/arm-cpus.in: Move information from fpu field of each
7250 cpu definition to the isa field.
7251 * config/arm/parsecpu.awk (fpu): Delete match rule.
7252 (gen_comm_data): Don't add bits from the CPU's FPU entry.
7254 2018-07-11 Richard Biener <rguenther@suse.de>
7257 * dwarf2out.c (init_sections_and_labels): Use
7258 output_asm_line_debug_info consistently.
7259 (dwarf2out_early_finish): Likewise.
7260 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
7263 2018-07-11 Richard Biener <rguenther@suse.de>
7265 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
7266 Rework father_bb setting in a way to avoid propagating constants
7267 multiple times on a loop body.
7269 2018-07-10 Mark Wielaard <mark@klomp.org>
7272 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
7274 2018-07-10 Richard Biener <rguenther@suse.de>
7276 * hash-map.h (hash_map::iterator::operator*): Return
7277 references to key and value.
7279 2018-07-10 Jakub Jelinek <jakub@redhat.com>
7282 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
7283 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
7284 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
7285 (gimplify_omp_for): For composite loops, move outer
7286 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
7287 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
7288 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
7289 TREE_LIST for both the original class iterator and the "last" helper
7290 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
7291 loop, remember has_decl_expr from outer composite loops for the
7292 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
7294 2018-07-09 Martin Sebor <msebor@redhat.com>
7298 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
7299 accessing implicitly initialized array elements.
7300 * expr.c (string_constant): Handle string initializers of
7301 character arrays within aggregates.
7302 * gimple-fold.c (fold_array_ctor_reference): Add argument.
7303 Store element offset. As a special case, handle zero size.
7304 (fold_nonarray_ctor_reference): Same.
7305 (fold_ctor_reference): Add argument. Store subobject offset.
7306 * gimple-fold.h (fold_ctor_reference): Add argument.
7308 2018-07-09 Paul Koning <ni1d@arrl.net>
7310 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
7311 (pdp11_insn_cost): New function.
7312 (pdp11_md_asm_adjust): New function.
7313 (TARGET_INVALID_WITHIN_DOLOOP): Define.
7314 (pdp11_rtx_costs): Update to match machine better.
7315 (output_addr_const_pdp11): Correct format mismatch warnings.
7316 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
7317 * config/pdp11/pdp11.md: General change to add base_cost and/or
7318 length attributes for use by new pdp11_insn_cost function.
7319 (MIN_BRANCH): Correct definition.
7321 (doloop_end): Use standard pattern name for looping pattern.
7322 (doloop_end_nocc): New.
7323 (movsf): Add another constraint alternative.
7324 (zero_extendqihi2): Add constraint alternatives for not in place
7326 (zero_extendhisi2): Remove.
7327 (shift patterns): Add CC handling variants.
7331 (define_peephole2): New peephole to recognize mov that sets CC for
7334 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7336 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
7337 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
7340 2018-07-09 Jakub Jelinek <jakub@redhat.com>
7343 * real.c (real_nextafter): Return true if result is denormal.
7345 2018-07-09 Martin Liska <mliska@suse.cz>
7347 * common.opt: Add back wrongly removed attribute.
7349 2018-07-09 Richard Biener <rguenther@suse.de>
7352 * dwarf2out.c (gen_block_die): For an early generated DIE
7353 always output high/low PC attributes.
7355 2018-07-09 Tom de Vries <tdevries@suse.de>
7357 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
7358 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
7359 onto VAR_DECL with abstract origin.
7361 2018-07-07 Jim Wilson <jimw@sifive.com>
7363 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
7365 2018-07-07 Tom de Vries <tdevries@suse.de>
7367 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
7369 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
7371 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
7373 (combine_bound): Use wide-int overflow calculation instead of
7375 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
7377 * fold-const.c (int_const_binop_2): Same.
7378 (extract_muldiv_1): Same.
7379 (fold_div_compare): Same.
7380 (fold_abs_const): Same.
7382 * poly-int.h (add): Same.
7386 * predict.c (predict_iv_comparison): Same.
7387 * profile-count.c (slow_safe_scale_64bit): Same.
7388 * simplify-rtx.c (simplify_const_binary_operation): Same.
7389 * tree-chrec.c (tree_fold_binomial): Same.
7390 * tree-data-ref.c (split_constant_offset_1): Same.
7391 * tree-if-conv.c (idx_within_array_bound): Same.
7392 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
7393 * tree-ssa-phiopt.c (minmax_replacement): Same.
7394 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
7395 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
7396 * vr-values.c (vr_values::adjust_range_with_scev): Same.
7397 * wide-int.cc (wi::add_large): Same.
7398 (wi::mul_internal): Same.
7399 (wi::sub_large): Same.
7400 (wi::divmod_internal): Same.
7401 * wide-int.h: Change overflow type to overflow_type for neg, add,
7402 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
7403 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
7404 mul_internal, divmod_internal.
7405 (overflow_type): New enum.
7406 (accumulate_overflow): New.
7408 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
7410 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
7411 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
7413 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
7415 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
7416 argument is checked for zero before entering loop, avoid checking again.
7418 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
7420 * gimplify.h (generic_expr_could_trap_p): Set as global function.
7421 * gimplify.h (generic_expr_could_trap_p): Likwise.
7422 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
7424 2018-07-06 Jakub Jelinek <jakub@redhat.com>
7426 PR tree-optimization/86401
7427 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
7428 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
7429 (fold_bit_and_mask): ... here. New helper function for match.pd.
7430 * fold-const.h (fold_bit_and_mask): Declare.
7431 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
7433 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
7436 * target.def (translate_mode_attribute): New hook.
7437 * targhooks.h (default_translate_mode_attribute): Declare.
7438 * targhooks.c (default_translate_mode_attribute): New function.
7439 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
7440 * doc/tm.texi: Regenerate.
7441 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
7442 (rs6000_translate_mode_attribute): New function.
7444 2018-07-06 Paul Koning <ni1d@arrl.net>
7446 * doc/md.texi (define_split): Document DONE and FAIL.
7447 (define_peephole2): Ditto.
7449 2018-07-05 Jeff Law <law@redhat.com>
7451 PR tree-optimization/86010
7452 * tree-ssa-dse.c (compute_trims): More aggressively trim at
7453 both the head and tail of mem* and str* calls.
7455 2018-07-05 Jim Wilson <jimw@sifive.com>
7457 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7459 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
7461 * config/aarch64/aarch64-simd.md: correct flags text for
7462 MIN_EXPR replacement.
7464 2018-07-05 James Clarke <jrtc27@jrtc27.com>
7466 * configure: Regenerated.
7468 2018-07-05 Carl Love <cel@us.ibm.com>
7470 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7471 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7472 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7473 VSX_BUILTIN_DOUBLEL_V4SF.
7475 2018-07-05 Martin Sebor <msebor@redhat.com>
7478 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7479 than its domain to compute its the upper bound of a char array.
7481 2018-07-05 Nathan Sidwell <nathan@acm.org>
7483 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7484 * doc/cpp.texi: Update comment.
7485 * doc/tm.texi: Rebuilt.
7486 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7487 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7488 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7489 headers affected by extern "C".
7490 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7491 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7492 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7493 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7494 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7495 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7496 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7497 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7498 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7499 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7500 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7501 config/sparc/sp64-elf.h, config/spu/spu.h,
7502 config/stormy16/stormy16.h, config/v850/v850.h,
7503 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7504 define NO_IMPLICIT_EXTERN_C.
7505 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7507 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7510 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7511 instead of GET_MODE_SIZE when comparing Units.
7513 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7516 * rtlanal.c (set_noop_p): Constrain on mode change,
7517 include hard-reg-set.h
7519 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7521 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7523 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7526 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7529 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7531 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7533 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7536 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7538 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7540 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7543 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7545 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7547 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7548 neg_*_op* variables.
7550 2018-07-04 Martin Liska <mliska@suse.cz>
7552 * tree-switch-conversion.c: Define
7553 max_ratio_for_speed and max_ratio_for_size constants.
7555 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7556 Martin Liska <mliska@suse.cz>
7561 * common.opt: Rename align options with 'str_' prefix.
7562 * common/config/i386/i386-common.c (set_malign_value): New
7564 (ix86_handle_option): Use it to set -falign-* options/
7565 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7566 type from int to string.
7567 * config/aarch64/aarch64.c: Update default values from int
7569 * config/alpha/alpha.c (alpha_override_options_after_change):
7571 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7572 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7573 max skip conditionally.
7574 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7575 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7576 max skip conditionally.
7577 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7578 max skip conditionally.
7579 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7580 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7581 max skip conditionally.
7582 * config/i386/i386.c (struct ptt): Change type from int to
7584 (ix86_default_align): Set default values.
7585 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7586 max skip conditionally.
7587 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7588 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7589 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7590 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7591 max skip conditionally.
7592 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7593 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7594 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7595 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7596 max skip conditionally.
7597 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7598 * config/ia64/ia64.c (ia64_option_override): Set default values
7599 for alignment options.
7600 * config/m68k/m68k.c: Handle new str_align_* options.
7601 * config/mips/mips.c (mips_set_compression_mode): Change
7603 (mips_option_override): Set default values for options.
7604 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7606 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7608 * config/rx/rx.c (rx_option_override): Likewise.
7609 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7610 (LABEL_ALIGN): Use align_labels_log.
7611 (LOOP_ALIGN): Use align_loops_align.
7612 * config/s390/s390.c (s390_asm_output_function_label): Use new
7614 * config/sh/sh.c (sh_override_options_after_change):
7615 Change type of constants.
7616 * config/spu/spu.c (spu_sched_init): Likewise.
7617 * config/sparc/sparc.c (sparc_option_override): Set default
7619 * config/visium/visium.c (visium_option_override): Likewise.
7620 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7621 emit p2align format with last argument if it's not needed.
7622 * doc/invoke.texi: Document extended format of -falign-*.
7623 * final.c: Use align_labels alignment.
7624 * flags.h (struct target_flag_state): Change type to use
7626 (struct align_flags_tuple): New.
7627 (struct align_flags): Likewise.
7628 (align_loops_log): Redefine macro to use new types.
7629 (align_loops_max_skip): Redefine macro to use new types.
7630 (align_jumps_log): Redefine macro to use new types.
7631 (align_jumps_max_skip): Redefine macro to use new types.
7632 (align_labels_log): Redefine macro to use new types.
7633 (align_labels_max_skip): Redefine macro to use new types.
7634 (align_functions_log): Redefine macro to use new types.
7635 (align_loops): Redefine macro to use new types.
7636 (align_jumps): Redefine macro to use new types.
7637 (align_labels): Redefine macro to use new types.
7638 (align_functions): Redefine macro to use new types.
7639 (align_functions_max_skip): Redefine macro to use new types.
7640 (align_loops_value): New macro.
7641 (align_jumps_value): New macro.
7642 (align_labels_value): New macro.
7643 (align_functions_value): New macro.
7644 * function.c (invoke_set_current_function_hook): Propagate
7645 alignment values from flags to global variables default in
7647 * ipa-icf.c (sem_function::equals_wpa): Use
7648 cl_optimization_option_eq instead of memcmp.
7649 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7651 (cl_optimization_stream_in): Likewise.
7652 * optc-save-gen.awk: Support strings in cl_optimization.
7653 * opth-gen.awk: Likewise.
7654 * opts.c (finish_options): Remove error checking of invalid
7656 (MAX_CODE_ALIGN): Remove.
7657 (MAX_CODE_ALIGN_VALUE): Likewise.
7658 (parse_and_check_align_values): New function.
7659 (check_alignment_argument): Likewise.
7660 (common_handle_option): Use check_alignment_argument.
7661 * opts.h (parse_and_check_align_values): Declare.
7662 * toplev.c (init_alignments): Remove.
7663 (read_log_maxskip): New.
7664 (parse_N_M): Likewise.
7665 (parse_alignment_opts): Likewise.
7666 (backend_init_target): Remove usage of init_alignments.
7667 * toplev.h (parse_alignment_opts): Declare.
7668 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7670 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7671 * tree.c (cl_option_hasher::equal): New.
7672 * varasm.c: Use new global macros.
7674 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7676 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7677 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7678 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7679 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7680 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7681 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7682 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7683 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7684 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7685 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7686 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7687 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7688 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7690 2018-07-04 Martin Liska <mliska@suse.cz>
7691 Jonathan Wakely <jwakely@redhat.com>
7693 * coverage.c: Use correct type.
7694 * doc/invoke.texi: Language correction.
7696 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
7699 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7700 ENDBRANCH for non-tail call which may return via indirect branch.
7701 * doc/extend.texi: Document indirect_return attribute.
7703 2018-07-03 Martin Sebor <msebor@redhat.com>
7705 PR tree-optimization/86274
7706 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7708 (format_floating): Correct handling of infinities and NaNs.
7710 2018-07-03 Martin Sebor <msebor@redhat.com>
7712 * print-tree.c (print_real_cst): New function.
7713 (print_node_brief): Call it.
7714 (print_node): Ditto.
7716 2018-07-03 Jeff Law <law@redhat.com>
7718 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7719 into a single pattern.
7721 * config/h8300/h8300.md (ors code_iterator): New.
7722 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7723 a single pattern and single splitter.
7724 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7725 (iorqi3_1, xorqi3_1): Likewise.
7726 (iorqi3, xorqi3 expanders): Similarly.
7728 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7729 (movmd_internal) into a single pattern using the P mode iterator.
7730 (movmd splitters): Similarly.
7731 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7732 (movsd splitters): Similarly.
7734 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7735 ADDB, ADDW and ADDL into a single ADD attribute which selects the
7736 right table based on the size of the operand.
7737 * config/h8300/h8300.md (length_table): Corresponding changes. All
7738 references to "addb", "addw" and "addl" changed to "add".
7739 (btst patterns): Merge two variants into a single pattern.
7740 (tstqi, tsthi): Likewise.
7741 (addhi3_incdec, addsi3_incdec): Likewise.
7742 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7743 (mulhi3, mulsi3): Likewise.
7744 (udivhi3, udivsi3): Likewise.
7745 (divhi3, divsi3): Likewise.
7746 (andorqi3, andorhi3, andorsi3): Likewise.
7748 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
7751 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7752 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7754 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7756 PR tree-optimization/85694
7757 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7758 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7760 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7761 (<u>avg<mode>3_ceil): New patterns.
7763 2018-07-03 David Malcolm <dmalcolm@redhat.com>
7765 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7766 scan-tree-dump directive.
7767 * gcc.dg/vect/slp-perm-2.c: Likewise.
7768 * gcc.dg/vect/slp-perm-3.c: Likewise.
7769 * gcc.dg/vect/slp-perm-5.c: Likewise.
7770 * gcc.dg/vect/slp-perm-6.c: Likewise.
7771 * gcc.dg/vect/slp-perm-7.c: Likewise.
7772 * gcc.dg/vect/slp-perm-8.c: Likewise.
7774 2018-07-03 Marek Polacek <polacek@redhat.com>
7777 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7779 2018-07-03 Richard Biener <rguenther@suse.de>
7782 * tree-ssa-structalias.c (find_func_clobbers): Properly
7783 handle indirect calls.
7785 2018-07-03 Jeff Law <law@redhat.com>
7787 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7788 (shifts): New code iterator.
7789 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7790 expander. Fix HImode handling on H8/SX.
7791 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7792 (subqi3, subhi3, subsi3 expanders): Likewise.
7793 (andqi3, andhi3, andsi3 expanders): Likewise.
7794 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7795 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7796 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7797 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7798 (zero_extendqihi2, zero_extendqisi2): Likewise.
7799 (extendqihi2, extendqisi2): Likewise.
7800 (rotlqi3, rotlhi3, rotlsi3): Likewise.
7801 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7802 (rotlqi3_1, rotlhi3_1): Likewise.
7803 (logicalhi3_sn, logicalsi3_sn): Likewise.
7804 (logicalhi3, logicalsi3): Likewise.
7806 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7808 * tree-vect-patterns.c (vect_recog_rotate_pattern)
7809 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7810 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7811 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7812 type to append_pattern_def_seq instead of creating a stmt_vec_info
7814 (build_mask_conversion): Likewise. Remove vinfo argument.
7815 (vect_add_conversion_to_patterm): Likewise, renaming to...
7816 (vect_add_conversion_to_pattern): ...this.
7817 (vect_recog_mask_conversion_pattern): Update call to
7818 build_mask_conversion. Pass the vector type to
7819 append_pattern_def_seq here too.
7820 (vect_recog_gather_scatter_pattern): Update call to
7821 vect_add_conversion_to_pattern.
7823 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7825 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7826 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7827 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7828 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7829 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7830 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7831 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7832 append_pattern_def_seq instead of new_pattern_def_seq.
7833 (vect_recog_divmod_pattern): Do both of the above.
7834 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7837 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7839 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7840 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7841 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7842 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7843 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7844 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7845 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7846 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7847 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7848 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7849 parameter with a single stmt_vec_info.
7850 (vect_recog_func_ptr): Likewise.
7851 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7852 (vect_try_gather_scatter_pattern): ...this.
7853 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7854 the stmt_vec_info of the statement to be matched. Don't clear
7855 STMT_VINFO_RELATED_STMT.
7856 (vect_pattern_recog): Update call accordingly.
7858 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7860 PR tree-optimization/85694
7861 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7862 (uavgM3_ceil): Document new optabs.
7863 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7864 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7866 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7867 (savg_ceil_optab): New optabs.
7868 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7869 (vect_vect_recog_func_ptrs): Add it.
7870 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7871 constant directly from the associated lhs.
7873 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7875 * tree-vect-patterns.c (vect_split_statement): New function.
7876 (vect_convert_input): Use it to try to split an existing cast.
7878 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7880 * poly-int.h (print_hex): New function.
7881 * dumpfile.h (dump_dec, dump_hex): Declare.
7882 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7883 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7884 min_input_precision, operation_precision and operation_sign.
7885 * tree-vect-patterns.c (vect_get_range_info): New function.
7886 (vect_same_loop_or_bb_p, vect_single_imm_use)
7887 (vect_operation_fits_smaller_type): Delete.
7888 (vect_look_through_possible_promotion): Add an optional
7889 single_use_p parameter.
7890 (vect_recog_over_widening_pattern): Rewrite to use new
7891 stmt_vec_info infomration. Handle one operation at a time.
7892 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7893 (vect_truncatable_operation_p, vect_set_operation_type)
7894 (vect_set_min_input_precision): New functions.
7895 (vect_determine_min_output_precision_1): Likewise.
7896 (vect_determine_min_output_precision): Likewise.
7897 (vect_determine_precisions_from_range): Likewise.
7898 (vect_determine_precisions_from_users): Likewise.
7899 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7900 (vect_vect_recog_func_ptrs): Put over_widening first.
7902 (vect_pattern_recog): Call vect_determine_precisions.
7904 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7906 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7907 statements that have been replaced by further pattern statements.
7908 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7910 2018-07-03 Richard Biener <rguenther@suse.de>
7912 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7913 always set *dt. Dump vectype in vectype overload.
7914 * dumpfile.h (dump_gimple_expr): New function.
7915 (dump_gimple_expr_loc): Likewise.
7916 * dumpfile.c (dump_gimple_expr): New function.
7917 (dump_gimple_expr_loc): Likewise.
7919 2018-07-02 Jeff Law <law@redhat.com>
7921 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7922 the H8/300, H8/300H and H8/S variants into a single pattern.
7923 (movhi_h8300, movqi_h8300hs): Similarly.
7924 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7925 (QHI mode iterator): New.
7927 * config/h8300/h8300.md: Remove trailing whitespace.
7929 2018-07-02 Jim Wilson <jimw@sifive.com>
7931 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7932 instead of emit_insn for interrupt returns.
7933 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7934 (riscv_sret, riscv_uret): Likewise.
7936 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7938 * pretty-print.c (selftest::test_pp_format): Move save and restore
7939 of quotes to class auto_fix_quotes, and add an instance.
7940 * selftest.c: Include "intl.h".
7941 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7942 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7943 * selftest.h (selftest::auto_fix_quotes): New class.
7945 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7947 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7948 (aarch64_sve_prepare_conditional_op): Remove.
7949 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7950 Allow aarch64_simd_reg_or_zero as select operand; remove
7951 the aarch64_sve_prepare_conditional_op call.
7952 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7953 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7954 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7955 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7956 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7957 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7958 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7959 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7960 and a splitters to match all of the *_any patterns.
7961 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7963 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7964 (SVE_COND_FP_BINARY_REV): Remove.
7965 (sve_int_op_rev, sve_fp_op_rev): New.
7966 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7967 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7968 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7969 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7970 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7971 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7972 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7973 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7974 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7976 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7977 Remove match_dup 1 from the inner unspec.
7978 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7980 * config/aarch64/aarch64.md (movprfx): New attr.
7981 (length): Default movprfx to 8.
7982 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7983 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7984 (*<su>mul<SVE_I>3_highpart): Likewise.
7985 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7986 (*v<ASHIFT><SVE_I>3): Likewise.
7987 (*<su><MAXMIN><SVE_I>3): Likewise.
7988 (*<su><MAXMIN><SVE_F>3): Likewise.
7989 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7990 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7991 (*div<SVE_F>4): Likewise.
7993 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7995 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7998 2018-07-02 Richard Biener <rguenther@suse.de>
8000 PR tree-optimization/86363
8001 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
8002 memset argument refers to a non-variable address.
8004 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8006 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
8007 {PLUS,MINUS}_EXPR code to...
8008 (adjust_symbolic_bound): ...here,
8009 (combine_bound): ...here,
8010 (set_value_range_with_overflow): ...and here.
8012 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8014 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
8016 (extract_range_from_abs_expr): ...here.
8018 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
8020 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
8021 -fno-omit-frame-pointer when not optimizing.
8023 2018-07-02 Martin Liska <mliska@suse.cz>
8026 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
8027 (propagate_nothrow): Likewise.
8029 2018-07-02 Martin Liska <mliska@suse.cz>
8032 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
8034 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8036 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
8037 function in r262149, changing "loc" param from source_location to
8038 const dump_location_t &.
8039 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
8040 declaration, as above.
8042 2018-07-01 Paul Koning <ni1d@arrl.net>
8044 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8045 -munit-asm, -mgnu-asm, -mdec-asm.
8046 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
8047 (pdp11_output_labelref): New.
8048 (pdp11_output_def): New.
8049 (pdp11_output_addr_vec_elt): New.
8050 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
8051 %# and %@ format codes.
8052 (pdp11_option_override): New.
8053 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
8054 (pdp11_output_ident): New.
8055 (pdp11_asm_named_section): New.
8056 (pdp11_asm_init_sections): New.
8057 (pdp11_file_start): New.
8058 (pdp11_file_end): New.
8059 (output_ascii): Use .ascii/.asciz for -mdec-asm.
8060 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
8061 %o, like %c but octal.
8062 (pdp11_option_override): New.
8063 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
8065 (DATA_SECTION_ASM_OP): Ditto.
8066 (READONLY_DATA_SECTION_ASM_OP): New.
8067 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
8068 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
8069 (ASM_OUTPUT_LABELREF): Ditto.
8070 (ASM_OUTPUT_DEF): Ditto.
8071 (ASM_OUTPUT_EXTERNAL): New.
8072 (ASM_OUTPUT_SOURCE_FILENAME): New.
8073 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
8074 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
8075 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
8076 %# and %@ format codes.
8077 * config/pdp11/pdp11.opt (mgnu-asm): New.
8078 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
8079 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
8080 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
8082 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
8084 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
8085 dereferencing path[] beyond its length.
8087 (debug_all_paths): New.
8088 (rewire_first_differing_edge): New.
8089 (adjust_paths_after_duplication): New.
8090 (duplicate_thread_path): Call adjust_paths_after_duplication.
8092 (thread_through_all_blocks): Add new argument to
8093 duplicate_thread_path.
8095 2018-06-30 Jim Wilson <jimw@sifive.com>
8097 * config/riscv/predicates.md (p2m1_shift_operand): New.
8098 (high_mask_shift_operand): New.
8099 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
8100 pattern using p2m1_shift_operand.
8101 (lshsi3_zero_extend_3+2): New combiner pattern using
8102 high_mask_shift_operand.
8104 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8106 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
8108 (vect_recog_rotate_pattern): ...here.
8109 (vect_convert_input): Try to insert casts of invariants in the
8111 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
8112 preheader to be empty.
8114 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8116 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
8117 vector type. If given, install it in the new statement's
8119 (vect_element_precision): New function.
8120 (vect_unpromoted_value): New struct.
8121 (vect_unpromoted_value::vect_unpromoted_value): New function.
8122 (vect_unpromoted_value::set_op): Likewise.
8123 (vect_look_through_possible_promotion): Likewise.
8124 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
8125 (vect_widened_op_tree, vect_convert_input): Likewise.
8126 (vect_convert_inputs, vect_convert_output): Likewise.
8127 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
8128 to handle the optional cast of the multiplication result and
8129 vect_widened_op_tree to detect the widened multiplication itself.
8130 Do not require the input and output of promotion casts to have
8131 the same sign, but base the signedness of the operation on the
8132 input rather than the result. If the pattern includes two
8133 promotions, check that those promotions have the same sign.
8134 Do not restrict the MULT_EXPR handling to a double-width result;
8135 handle quadruple-width results and wider. Use vect_convert_inputs
8136 to convert the inputs to the common type.
8137 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
8138 to handle the optional cast of the ABS result. Also allow a sign
8139 change or a sign extension between the ABS and MINUS.
8140 Use vect_widened_op_tree to detect the widened subtraction and use
8141 vect_convert_inputs to convert the inputs to the common type.
8142 (vect_handle_widen_op_by_const): Delete.
8143 (vect_recog_widen_op_pattern): New function.
8144 (vect_recog_widen_mult_pattern): Use it.
8145 (vect_recog_widen_shift_pattern): Likewise.
8146 (vect_recog_widen_sum_pattern): Use
8147 vect_look_through_possible_promotion to handle the promoted
8150 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8152 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
8153 the containing gimple_seq *.
8154 * gimple-iterator.h (gsi_for_stmt): Declare it.
8155 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
8156 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
8157 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8158 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8159 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
8161 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
8163 (vect_mark_pattern_stmts): ...here. Handle cases in which the
8164 statement being replaced is part of an existing pattern
8165 definition sequence, inserting the new pattern statements before
8167 (vect_pattern_recog_1): Don't return a bool. If the statement
8168 is already part of a pattern, instead apply pattern matching
8169 to the pattern definition statements. Don't clear the
8170 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
8171 (vect_pattern_recog): Don't break after the first match;
8172 continue processing the pattern definition statements instead.
8173 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
8175 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8177 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8178 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8179 (vect_recog_widen_sum_pattern): Use it.
8181 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8183 * tree-vect-loop.c (vectorizable_reduction): Assert that the
8184 phi is not a pattern statement and has not been replaced by
8185 a pattern statement.
8186 * tree-vect-patterns.c (type_conversion_p): Don't check
8187 STMT_VINFO_IN_PATTERN_P.
8188 (vect_recog_vector_vector_shift_pattern): Likewise.
8189 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
8190 the pattern statement rather than the original statement; check
8191 directly for a WIDEN_MULT_EXPR here.
8192 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
8193 vect_is_simple_use to return the pattern statement rather
8194 than the original statement; use is_pattern_stmt_p to check
8195 for such a pattern statement.
8196 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
8197 to return the pattern statement rather than the original statement;
8198 don't do the same transformation here.
8199 (vect_is_simple_use): If the defining statement has been replaced
8200 by a pattern statement, return the pattern statement instead.
8201 Remove the corresponding (local) transformation from the vectype
8204 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8206 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
8207 end and default to null.
8208 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8209 (vectorizable_reduction): Update calls accordingly, dropping the
8210 gimple ** argument if the passed-back statement isn't needed.
8211 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
8212 (vect_recog_rotate_pattern): Likewise.
8213 (vect_recog_mask_conversion_pattern): Likewise.
8214 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8215 (vect_mask_constant_operand_p): Likewise.
8216 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
8217 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
8218 (get_group_load_store_type, get_load_store_type): Likewise.
8219 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
8220 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
8221 (vectorizable_conversion, vectorizable_assignment): Likewise.
8222 (vectorizable_shift, vectorizable_operation): Likewise.
8223 (vectorizable_store, vect_is_simple_cond): Likewise.
8224 (vectorizable_condition, vectorizable_comparison): Likewise.
8225 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
8226 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
8227 and move it to the end. Cope with null def_stmt_outs.
8229 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8231 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
8233 2018-06-29 Jeff Law <law@redhat.com>
8235 * config/v850/v850.c (v850_legitimate_address_p): Handle large
8236 displacements for TARGET_V850E2V3 and newer.
8237 (TARGET_LRA_P): Remove. Defaults to LRA now.
8238 * config/v850/v850.md (sign23byte_load): Remove.
8239 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
8240 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
8242 2018-06-29 Martin Liska <mliska@suse.cz>
8245 * coverage.c (coverage_init): Mangle full path name.
8246 * doc/invoke.texi: Document the change.
8247 * gcov-io.c (mangle_path): New.
8248 * gcov-io.h (mangle_path): Likewise.
8249 * gcov.c (mangle_name): Use mangle_path for path mangling.
8251 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8253 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
8254 if starting source register is not even.
8256 2018-06-29 Martin Liska <mliska@suse.cz>
8258 PR tree-optimization/86263
8259 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
8260 Make edge redirection.
8262 2018-06-29 David Malcolm <dmalcolm@redhat.com>
8264 * dumpfile.c (dump_loc): Add indentation based on scope depth.
8265 (dump_scope_depth): New variable.
8266 (get_dump_scope_depth): New function.
8267 (dump_begin_scope): New function.
8268 (dump_end_scope): New function.
8269 * dumpfile.h (get_dump_scope_depth): New declaration.
8270 (dump_begin_scope): New declaration.
8271 (dump_end_scope): New declaration.
8272 (class auto_dump_scope): New class.
8273 (AUTO_DUMP_SCOPE): New macro.
8274 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
8277 2018-06-29 Richard Biener <rguenther@suse.de>
8279 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
8280 compute_all_dependences succeeds.
8281 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
8282 exceed --param loop-max-datarefs-for-datadeps.
8284 2018-06-29 Jakub Jelinek <jakub@redhat.com>
8286 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
8288 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
8291 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
8292 alternative 0 in preferred_for_speed attribute.
8294 2018-06-28 Paul Koning <ni1d@arrl.net>
8296 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
8297 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
8298 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
8299 * config/pdp11/pdp11.md: Correct "length" attribute calculation
8300 for shift insn patterns.
8302 2018-06-28 David Malcolm <dmalcolm@redhat.com>
8304 * cgraph.c (cgraph_node::get_body): Replace assignments to
8305 "dump_file" with calls to set_dump_file.
8306 * dumpfile.c (alt_dump_file): Make static, and group with...
8307 (alt_flags): ...this definition.
8308 (dumps_are_enabled): New variable.
8309 (refresh_dumps_are_enabled): New function.
8310 (set_dump_file): New function.
8311 (set_alt_dump_file): New function.
8312 (gcc::dump_manager::dump_start): Replace assignments to
8313 "dump_file" and "alt_dump_file" with calls to set_dump_file and
8315 (gcc::dump_manager::dump_finish): Likewise.
8316 * dumpfile.h (alt_dump_file): Delete decl.
8317 (dumps_are_enabled): New variable decl.
8318 (set_dump_file): New function decl.
8319 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
8321 * tree-nested.c (lower_nested_functions): Replace assignments to
8322 "dump_file" with calls to set_dump_file.
8324 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
8326 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
8327 goto_locus of each outgoing edge of each basic block.
8329 2018-06-28 Richard Biener <rguenther@suse.de>
8331 * dwarf2out.c (decl_scope_table): Remove.
8332 (push_decl_scope): Likewise.
8333 (pop_decl_scope): Likewise.
8334 (gen_type_die_for_member): Do not call push/pop_decl_scope.
8335 (gen_struct_or_union_type_die): Likewise.
8336 (gen_tagged_type_die): Likewise.
8337 (dwarf2out_init): Do not initialize decl_scope_table.
8338 (dwarf2out_c_finalize): Do not free it.
8340 2018-06-28 Richard Biener <rguenther@suse.de>
8342 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
8343 deciding whether to not re-use a DIE.
8345 2018-06-28 Richard Biener <rguenther@suse.de>
8347 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
8348 DW_AT_abstract_origin attribute.
8350 2018-06-28 Martin Liska <mliska@suse.cz>
8352 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8353 Use newly introduced constants.
8354 * tree-switch-conversion.h (struct jump_table_cluster):
8355 Define max_ratio_for_size and max_ratio_for_speed.
8357 2018-06-28 Martin Liska <mliska@suse.cz>
8359 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8360 Add new checking assert to catch invalid state.
8361 (jump_table_cluster::can_be_handled): Handle single case
8363 (jump_table_cluster::is_beneficial): Bail out for such case.
8364 (bit_test_cluster::find_bit_tests):
8365 Add new checking assert to catch invalid state.
8366 (bit_test_cluster::can_be_handled): Handle single case
8368 (bit_test_cluster::is_beneficial): Bail out for such case.
8369 (switch_decision_tree::analyze_switch_statement):
8372 2018-06-28 Martin Liska <mliska@suse.cz>
8374 * common.opt: Introduce -completion option.
8375 * gcc.c (driver_handle_option): Handle it.
8376 (driver::main): Print completions if completion
8378 * opt-suggestions.c (option_proposer::get_completions):
8380 (option_proposer::suggest_completion): Likewise.
8381 (option_proposer::find_param_completions): Likewise.
8382 (verify_autocompletions): Likewise.
8383 (test_completion_valid_options): Likewise.
8384 (test_completion_valid_params): Likewise.
8385 (in_completion_p): Likewise.
8386 (empty_completion_p): Likewise.
8387 (test_completion_partial_match): Likewise.
8388 (test_completion_garbage): Likewise.
8389 (opt_proposer_c_tests): Likewise.
8390 * opt-suggestions.h: Declare new functions.
8391 * opts.c (common_handle_option): Handle OPT__completion_.
8392 * selftest-run-tests.c (selftest::run_tests): Add
8393 opt_proposer_c_tests.
8394 * selftest.c (assert_str_startswith): New.
8395 * selftest.h (assert_str_startswith): Likewise.
8396 (opt_proposer_c_tests): New.
8397 (ASSERT_STR_STARTSWITH): Likewise.
8399 2018-06-28 Martin Liska <mliska@suse.cz>
8401 * Makefile.in: Add opt-suggestions.o.
8402 * gcc-main.c: Include opt-suggestions.h.
8403 * gcc.c (driver::driver): Likewise.
8404 (driver::~driver): Remove m_option_suggestions.
8405 (driver::build_option_suggestions): Moved to option_proposer.
8406 (driver::suggest_option): Likewise.
8407 (driver::handle_unrecognized_options): Use option_proposer.
8408 * gcc.h (class driver): Add new memver m_option_proposer.
8409 * opt-suggestions.c: New file.
8410 * opt-suggestions.h: New file.
8412 2018-06-28 Martin Liska <mliska@suse.cz>
8414 * vec.h (class auto_string_vec): New (moved from auto_argvec).
8415 (auto_string_vec::~auto_string_vec): Likewise.
8417 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
8419 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
8420 prevent_decl_creation_for_types fields up and add reset_location field.
8421 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
8422 statement if id->reset_location is true.
8423 (copy_edges_for_bb): Do not set goto_locus on the new edges if
8424 id->reset_location is true.
8425 (copy_phis_for_bb): Force input_location on the arguments if
8426 id->reset_location is true.
8427 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
8428 is set on the function to be inlined.
8430 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
8432 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
8434 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
8436 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
8437 registers for Pmode.
8438 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
8439 hard registers for the clobbered pseudo.
8441 2018-06-27 Paul Koning <ni1d@arrl.net>
8443 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8444 mutually exclusive options.
8445 * config/pdp11/constraints.md (h): New constraint.
8446 (O): Update definition to match shift code generation.
8447 (D): New constraint.
8448 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8450 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8452 (output_jump): Change arguments.
8453 (pdp11_fixed_cc_regs): New function.
8454 (pdp11_cc_mode): Ditto.
8455 (pdp11_expand_shift): Ditto.
8456 (pdp11_assemble_shift): Ditto.
8457 (pdp11_small_shift): Ditto.
8458 (pdp11_branch_cost): Remove.
8459 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8461 (pdp11_register_move_cost): Update for CC registers.
8462 (pdp11_rtx_costs): Add case for LSHIFTRT.
8463 (pdp11_output_jump): Add CCNZ mode conditional branches.
8464 (notice_update_cc_on_set): Remove.
8465 (pdp11_cc_mode): New function.
8466 (simple_memory_operand): Correct pre/post decrement case.
8467 (no_side_effect_operand): New function.
8468 (pdp11_regno_reg_class): Add CC_REGS class.
8469 (pdp11_fixed_cc_regs): New function.
8470 (pdp11_small_shift): New function.
8471 (pdp11_expand_shift): New function to expand shift insns.
8472 (pdp11_assemble_shift): New function to output shifts.
8473 (pdp11_branch_cost): Remove.
8474 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8475 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8476 (WCHAR_TYPE): Ditto.
8477 (PTRDIFF_TYPE): Ditto.
8478 (ADJUST_INSN_LENGTH): New macro.
8479 (FIXED_REGISTERS): Add CC registers.
8480 (CALL_USED_REGISTERS): Ditto.
8482 (REG_CLASS_NAMES): Ditto.
8483 (REG_CLASS_CONTENTS): Ditto.
8484 (SELECT_CC_MODE): Use new function.
8485 (TARGET_FLAGS_REGNUM): New macro.
8486 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8487 (cc0_reg_rtx): Remove.
8488 (CC_STATUS_MDEP): Remove.
8489 (CC_STATUS_MDEFP_INIT): Remove.
8490 (CC_IN_FPU): Remove.
8491 (NOTICE_UPDATE_CC): Remove.
8492 (REGISTER_NAMES): Add CC registers.
8493 (BRANCH_COST): Change to constant 1.
8494 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8496 * config/pdp11/pdp11.opt (mbcopy): Remove.
8497 (mbcopy-builtin): Remove.
8498 (mbranch-cheap): Remove.
8499 (mbranch-expensive): Remove.
8500 * config/pdp11/predicates.md (expand_shift_operand): Update to
8501 match shift code generation.
8502 (ccnz_operator): New predicate.
8503 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8504 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8505 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8507 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8508 description of O constraint.
8510 2018-06-27 Jeff Law <law@redhat.com>
8511 Austin Law <austinklaw@gmail.com>
8513 * config/v850/v850.md (addsi3_set_flags): New pattern.
8514 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8515 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8516 (zero_extendhisi2_v850_set_flags): Likewise.
8517 (zero_extendqisi2_v850_set_flags): Likewise.
8518 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8519 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8520 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8522 * config/v850/v850-protos.h (notice_update_cc): Remove.
8523 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8524 (v850_print_operand): Handle 'D' and "d".
8525 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8526 Add handling of arithmetic/logical operations compared against zero.
8527 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8528 Do not look at v850_compare_op, instead get mode from last argument.
8529 (v850_gen_compare): Remove
8530 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8531 after reload for prologue insns.
8532 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8534 (construct_save_jarl): Likewise.
8535 (TARGET_FLAGS_REGNUM): Define.
8536 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8537 (NOTICE_UPDATE_CC): Remove.
8538 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8539 than cc0. Conditionalize on reload_completed.
8540 (cmpsi_insn, setfcc_insn): Likewise.
8541 (tst1 splitter): Turn into define_and_split which sets the flags
8543 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8544 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8545 (cstoresf4, cstoredf4): Clobber the flags.
8546 (cmpsi, cmpsf, cmpdf): Remove expanders.
8547 (setf_insn): Remove pattern.
8548 (addsi3): Turn into define_and_split which clobbers the flags after
8549 reload and a suitable pattern (addsi3_clobber_flags) for use after
8551 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8552 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8553 (ashrsi3, ashrsi3_v850e2): Likewise.
8554 (bins): Clobber the flags.
8555 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8556 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8557 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8558 (call_value_internal_short, call_value_internal_long): Likewise.
8559 (callt_save_interrupt, callt_return_interrupt): Likewise.
8560 (save_interrupt, return_interrupt): Likewise.
8561 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8562 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8563 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8564 (All FP comparisons): Only allow after reload has completed.
8566 (divh, divhu): Tweak output template.
8567 (branch_z_normal, branch_z_invert): Remove
8568 (branch_nz_normal, branch_nz_invert): Likewise.
8569 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8571 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8572 * config/v850/v850.c (notice_update_cc): Remove.
8573 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8574 (CC_NO_CARRY): Likewise.
8575 (NOTICE_UPDATE_CC): Define to nothing.
8576 * config/v850/v850.md: Remove block comment on cc0 handling
8577 Remove "cc" attribute from all patterns. Remove cc_status handling
8578 from all patterns. Minor formatting fixes.
8580 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8582 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8583 (cortex-a76.cortex-a55): Likewise.
8584 * config/aarch64/aarch64-tune.md: Regenerate.
8585 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8586 cortex-a76.cortex-a55.
8588 2018-06-27 Jeff Law <law@redhat.com>
8590 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8591 (MULTILIB_DIRNAMES): Similarly.
8593 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8595 * gimple.h (gimple_return_retbnd): Delete.
8596 (gimple_return_set_retbnd): Likewise.
8597 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8598 gimple_return_set_retbnd.
8599 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8600 gimple_return_retbnd and adjust.
8601 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8602 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8603 Explicitly return NULL in a couple more cases. Move assertion
8604 on debug statements and remove unreachable code.
8605 (reset_debug_binding): Do not test id->retbnd.
8606 (expand_call_inline): Do not set it.
8608 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8610 * configure.ac: Add --disable-gcov option.
8611 * configure: Regenerate.
8612 * Makefile.in: Honour @enable_gcov@.
8613 * doc/install.texi: Document --disable-gcov.
8615 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8617 * config/arm/arm-cpus.in (cortex-a76): New entry.
8618 (cortex-a76.cortex-a55): Likewise.
8619 * config/arm/arm-tables.opt: Regenerate.
8620 * config/arm/arm-tune.md: Likewise.
8621 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8622 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8623 cortex-a76.cortex-a55.
8625 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8628 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8630 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8632 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8634 (EPILOGUE_USES): Likewise.
8636 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8638 * tree-inline.c (remap_location): New function extracted from...
8639 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8640 (copy_phis_for_bb): ...here. Call remap_location.
8641 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8643 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8645 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8646 unaligned vsx for 16B memset.
8648 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8651 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8652 ieee128_float_type_node to long_double_type_node unless
8653 TARGET_LONG_DOUBLE_128 is set.
8655 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8657 * cfgloop.c (get_loop_location): Convert return type from
8658 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8659 by implicit construction from rtx_insn *, and using
8660 dump_user_location_t::from_function_decl for the fallback case.
8661 * cfgloop.h (get_loop_location): Convert return type from
8662 location_t to dump_user_location_t.
8663 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8664 dump_printf_loc to pass in a dump_location_t rather than a
8665 location_t, via the gimple stmt.
8666 * coverage.c (get_coverage_counts): Update calls to
8667 dump_printf_loc to pass in dump_location_t rather than a
8669 * doc/optinfo.texi (Dump types): Convert example of
8670 dump_printf_loc from taking "locus" to taking "insn". Update
8671 description of the "_loc" calls to cover dump_location_t.
8672 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8674 (dump_user_location_t::dump_user_location_t): New constructors,
8675 from gimple *stmt and rtx_insn *.
8676 (dump_user_location_t::from_function_decl): New function.
8677 (dump_loc): Make static.
8678 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8679 const dump_location_t &.
8680 (dump_generic_expr_loc): Delete.
8681 (dump_printf_loc): Convert param "loc" from location_t to
8682 const dump_location_t &.
8683 (selftest::test_impl_location): New function.
8684 (selftest::dumpfile_c_tests): New function.
8685 * dumpfile.h: Include "profile-count.h".
8686 (class dump_user_location_t): New class.
8687 (struct dump_impl_location_t): New struct.
8688 (class dump_location_t): New class.
8689 (dump_printf_loc): Convert 2nd param from source_location to
8690 const dump_location_t &.
8691 (dump_generic_expr_loc): Delete.
8692 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8693 const dump_location_t &.
8694 * gimple-fold.c (fold_gimple_assign): Update call to
8695 dump_printf_loc to pass in a dump_location_t rather than a
8696 location_t, via the gimple stmt.
8697 (gimple_fold_call): Likewise.
8698 * gimple-loop-interchange.cc
8699 (loop_cand::analyze_iloop_reduction_var): Update for change to
8700 check_reduction_path.
8701 (tree_loop_interchange::interchange): Update for change to
8703 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8704 change in return-type of find_loop_location.
8705 (graphite_regenerate_ast_isl): Likewise.
8706 * graphite-optimize-isl.c (optimize_isl): Likewise.
8707 * graphite.c (graphite_transform_loops): Likewise.
8708 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8709 pass in a dump_location_t rather than a location_t, via the
8711 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8712 * ipa.c (walk_polymorphic_call_targets): Likewise.
8713 * loop-unroll.c (report_unroll): Convert "locus" param from
8714 location_t to dump_location_t.
8715 (decide_unrolling): Update for change to get_loop_location's
8717 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8718 location_t to dump_user_location_t.
8719 (grid_find_single_omp_among_assignments_1): Updates calls to
8720 dump_printf_loc to pass in a dump_location_t rather than a
8721 location_t, via the gimple stmt.
8722 (grid_parallel_clauses_gridifiable): Convert "tloc" from
8723 location_t to dump_location_t. Updates calls to dump_printf_loc
8724 to pass in a dump_location_t rather than a location_t, via the
8726 (grid_inner_loop_gridifiable_p): Likewise.
8727 (grid_dist_follows_simple_pattern): Likewise.
8728 (grid_gfor_follows_tiling_pattern): Likewise.
8729 (grid_target_follows_gridifiable_pattern): Likewise.
8730 (grid_attempt_target_gridification): Convert initialization
8731 of local "grid" from memset to zero-initialization; FIXME: does
8732 this require C++11? Update call to dump_printf_loc to pass in a
8733 optinfo_location rather than a location_t, via the gimple stmt.
8734 * profile.c (read_profile_edge_counts): Updates call to
8735 dump_printf_loc to pass in a dump_location_t rather than a
8737 (compute_branch_probabilities): Likewise.
8738 * selftest-run-tests.c (selftest::run_tests): Call
8740 * selftest.h (dumpfile_c_tests): New decl.
8741 * tree-loop-distribution.c (pass_loop_distribution::execute):
8742 Update for change in return type of find_loop_location.
8743 * tree-parloops.c (parallelize_loops): Likewise.
8744 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8745 "locus" from location_t to dump_user_location_t.
8746 (canonicalize_loop_induction_variables): Likewise.
8747 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8748 for change in return type of find_loop_location.
8749 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8750 to dump_printf_loc to pass in a dump_location_t rather than a
8751 location_t, via the stmt.
8752 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8754 * tree-vect-loop-manip.c (find_loop_location): Convert return
8755 type from source_location to dump_user_location_t.
8756 (vect_do_peeling): Update for above change.
8757 (vect_loop_versioning): Update for change in type of
8759 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8760 from location_t to dump_user_location_t.
8761 (vect_estimate_min_profitable_iters): Update for change in type
8763 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8764 location_t to dump_location_t.
8765 (vect_slp_bb): Update for change in type of vect_location.
8766 * tree-vectorizer.c (vect_location): Convert from source_location
8767 to dump_user_location_t.
8768 (try_vectorize_loop_1): Update for change in vect_location's type.
8769 (vectorize_loops): Likewise.
8770 (increase_alignment): Likewise.
8771 * tree-vectorizer.h (vect_location): Convert from source_location
8772 to dump_user_location_t.
8773 (find_loop_location): Convert return type from source_location to
8774 dump_user_location_t.
8775 (check_reduction_path): Convert 1st param from location_t to
8776 dump_user_location_t.
8777 * value-prof.c (check_counter): Update call to dump_printf_loc to
8778 pass in a dump_user_location_t rather than a location_t; update
8779 call to error_at for change in type of "locus".
8780 (check_ic_target): Update call to dump_printf_loc to
8781 pass in a dump_user_location_t rather than a location_t, via the
8784 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
8786 * config/s390/s390.h (enum processor_flags): Do not use
8787 default tune parameter when -march was specified.
8789 2018-06-26 Jakub Jelinek <jakub@redhat.com>
8792 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8793 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8796 2018-06-26 Richard Biener <rguenther@suse.de>
8798 PR tree-optimization/86287
8800 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8801 (vect_analyze_loop): Initialize n_stmts.
8803 2018-06-26 Richard Biener <rguenther@suse.de>
8806 * fold-const.c (fold_convertible_p): Pointer extension
8809 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8812 * dwarf2out.c (loc_list_has_views): Adjust comments.
8813 (dw_loc_list): Split single cross-partition range with
8816 2018-06-25 Jeff Law <law@redhat.com>
8818 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8819 on -mbig-switch by default.
8821 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8823 (const_float_0_operand): Remove unused predicate.
8824 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8825 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8826 (recipsf2): New expander. Original pattern now called
8828 (recipdf2, recipdf2_insn): Similarly.
8829 (rsqrtsf2, rsqrtsf2_insn): Similarly
8830 (rsqrtdf2, rsqrtdf2_insn): Similarly
8832 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8834 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8835 Simplify logic for FreeBSD (twice).
8837 2018-06-25 Martin Sebor <msebor@redhat.com>
8839 PR tree-optimization/86204
8840 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8841 a strnlen result if it's less than the length of the string.
8843 2018-06-25 Martin Sebor <msebor@redhat.com>
8845 PR tree-optimization/85700
8846 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8847 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8848 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8850 2018-06-25 Martin Sebor <msebor@redhat.com>
8852 * doc/extend.texi (Zero-length arrays): Update and clarify.
8854 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8856 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8857 added IEEE/IBM long double multilib support on PowerPC little
8858 endian Linux systems.
8859 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8860 (MULTILIB_DEFAULTS): Likewise.
8861 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8863 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8864 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8865 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8867 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8870 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8871 (REORDER_45): Likewise.
8873 2018-06-25 Jeff Law <law@redhat.com>
8875 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8876 dividend to 32 bits. Adjust length.
8877 (udivmodhi4): Cleanup output template. Fix length.
8879 2018-06-25 Carl Love <cel@us.ibm.com>
8881 * config/rs6000/vsx.md: Change word selector to prefered location.
8883 2018-06-25 Richard Biener <rguenther@suse.de>
8885 PR tree-optimization/86304
8886 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8887 epilogue-if-converted loops as well.
8889 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8891 * lto-section-out.c (lto_begin_section): Do not print section
8892 name for noaddr and unnumbered dumps.
8894 2018-06-25 Richard Biener <rguenther@suse.de>
8896 * tree-vectorizer.h (struct vec_info_shared): New structure
8897 with parts split out from struct vec_info and loop_nest from
8898 struct _loop_vec_info.
8899 (struct vec_info): Adjust accordingly.
8900 (struct _loop_vec_info): Likewise.
8901 (LOOP_VINFO_LOOP_NEST): Adjust.
8902 (LOOP_VINFO_DATAREFS): Likewise.
8903 (LOOP_VINFO_DDRS): Likewise.
8904 (struct _bb_vec_info): Likewise.
8905 (BB_VINFO_DATAREFS): Likewise.
8906 (BB_VINFO_DDRS): Likewise.
8907 (struct _stmt_vec_info): Add dr_aux member.
8908 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8909 (DR_MISALIGNMENT_UNINITIALIZED): New.
8910 (set_dr_misalignment): Adjust.
8911 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8912 (vect_analyze_loop): Adjust prototype.
8913 (vect_analyze_loop_form): Likewise.
8914 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8915 Compute dependences lazily.
8916 (vect_record_base_alignments): Use shared datarefs/ddrs.
8917 (vect_verify_datarefs_alignment): Likewise.
8918 (vect_analyze_data_refs_alignment): Likewise.
8919 (vect_analyze_data_ref_accesses): Likewise.
8920 (vect_analyze_data_refs): Likewise.
8921 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8922 constructor parameter for shared part.
8923 (vect_analyze_loop_form): Pass in shared part and adjust.
8924 (vect_analyze_loop_2): Pass in storage for the number of
8925 stmts. Move loop nest finding to the caller. Compute
8927 (vect_analyze_loop): Pass in shared part.
8928 (vect_transform_loop): Verify shared datarefs are unchanged.
8929 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8930 constructor parameter for shared part.
8931 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8932 (vect_slp_bb): Verify shared datarefs are unchanged before
8934 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8936 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8937 DR_MISALIGNMENT_UNINITIALIZED.
8938 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8939 parameter for shared part.
8940 (vec_info::~vec_info): Adjust.
8941 (vec_info_shared::vec_info_shared): New.
8942 (vec_info_shared::~vec_info_shared): Likewise.
8943 (vec_info_shared::save_datarefs): Likewise.
8944 (vec_info_shared::check_datarefs): Likewise.
8945 (try_vectorize_loop_1): Construct shared part live for analyses
8946 of a single loop for multiple vector sizes.
8947 * tree-parloops.c (gather_scalar_reductions): Adjust.
8949 2018-06-25 Richard Biener <rguenther@suse.de>
8951 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8952 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8953 (vect_analyze_data_refs): Remove similar code from here and
8954 simplify accordingly.
8956 2018-06-25 Richard Biener <rguenther@suse.de>
8958 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8959 for reverse storage order accesses rather than asserting
8960 they cannot happen here.
8962 2018-06-25 Tom de Vries <tdevries@suse.de>
8965 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8966 Use data16 instead of .byte for insn prefix.
8968 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8971 * parser.c (make_char_string_pack): Pass this literal chars
8972 through cpp_interpret_string.
8973 (cp_parser_userdef_numeric_literal): Check the result of
8974 make_char_string_pack.
8976 2018-06-24 Maya Rashish <coypu@sdf.org>
8978 * ginclude/stddef.h: Simplify conditions around avoiding
8979 re-definition of __size_t.
8981 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8983 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8984 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8986 2018-06-22 Maya Rashish <coypu@sdf.org>
8988 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8989 made to OpenBSD and NetBSD through the years.
8991 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8993 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8994 behavior of vec_pack (vector double, vector double) to match
8995 behavior of vec_float2 (vector double, vector double).
8997 2018-06-22 Olivier Hainque <hainque@adacore.com>
8999 * gimplify.c (gimplify_function_tree): Prevent creation
9000 of a trampoline for the address of the current function
9001 passed to entry/exit instrumentation hooks.
9003 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
9006 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
9009 2018-06-22 Martin Liska <mliska@suse.cz>
9011 PR tree-optimization/86263
9012 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9013 Bail out if is_enabled is false.
9014 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
9016 (jump_table_cluster::is_enabled): New function.
9018 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9020 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9021 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
9022 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
9023 (lto_input_ts_binfo_tree_pointers): Likewise.
9024 * tree-streamer-out.c (streamer_write_tree_bitfields,
9025 write_ts_binfo_tree_pointers): Likewise.
9026 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
9028 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9030 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
9032 2018-06-22 Martin Liska <mliska@suse.cz>
9034 * symbol-summary.h (get): Make it pure and inline move
9035 functionality from ::get function.
9036 (get): Remove and inline into ::get and ::get_create.
9037 (get_create): Move code from ::get function.
9039 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9042 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
9043 -x assembler-with-cpp.
9045 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9047 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
9048 _FILE_OFFSET_BITS=64 for C++.
9050 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
9052 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
9053 conversion insn that shows up when pr85657-3.c is compiled using
9054 IEEE 128-bit long double.
9055 (neg<mode>2_internal): Use the correct mode to check whether the
9056 mode is IBM extended.
9057 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
9058 multiply and divide external functions from being created more
9061 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9063 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
9065 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
9066 the edge can be forwarded.
9067 (cfg_layout_merge_blocks): Likewise.
9069 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9071 * except.c (finish_eh_generation): Commit edge insertions only after
9072 the EH edges have been redirected from post-landing to landing pads.
9074 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9076 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
9077 create_tmp_var_for to create the FRAME decl.
9078 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
9080 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9082 * tree-inline.c (copy_edges_for_bb): Minor tweak.
9083 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
9084 debug statement when resetting its value.
9085 (expand_call_inline): Copy the locus of the call onto the assignment
9086 of the return value, if any. Use local variable in more cases.
9088 2018-06-21 Martin Liska <mliska@suse.cz>
9090 * ipa-pure-const.c (propagate_nothrow): Use
9091 funct_state_summaries->get.
9092 (dump_malloc_lattice): Likewise.
9093 (propagate_malloc): Likewise.
9095 2018-06-21 Richard Biener <rguenther@suse.de>
9097 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
9098 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
9099 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
9101 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
9102 BLOCK_ABSTRACT_ORIGIN unconditionally.
9104 2018-06-21 David Malcolm <dmalcolm@redhat.com>
9106 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
9108 * ipa-reference.c (ipa_reference_c_finalize): Delete
9109 ipa_ref_opt_sum_summaries and set it to NULL.
9111 2018-06-21 Tom de Vries <tdevries@suse.de>
9113 PR tree-optimization/85859
9114 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
9115 test with comment from bb_no_side_effects_p.
9117 2018-06-21 Richard Biener <rguenther@suse.de>
9119 PR tree-optimization/86232
9120 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
9121 max for constant niter.
9123 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9125 * config/aarch64/aarch64-simd.md
9126 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
9128 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9130 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
9131 Make opernads of the unspec commutative.
9133 2018-06-21 Richard Biener <rguenther@suse.de>
9135 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
9136 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
9137 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
9138 (vect_analyze_data_ref_dependence): Re-order checks to deal with
9140 (vect_record_base_alignments): Do not record base alignment
9141 for gathers or scatters.
9142 (vect_compute_data_ref_alignment): Drop return value that is always
9143 true. Bail out early for gathers or scatters.
9144 (vect_enhance_data_refs_alignment): Bail out early for gathers
9146 (vect_find_same_alignment_drs): Likewise.
9147 (vect_analyze_data_refs_alignment): Remove dead code.
9148 (vect_slp_analyze_and_verify_node_alignment): Likewise.
9149 (vect_analyze_data_refs): For possible gathers or scatters do
9150 not create an alternate DR, just check their possible validity
9151 and mark them. Adjust DECL_NONALIASED handling to not rely
9153 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
9154 update inits of gathers or scatters.
9155 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
9156 Also copy gather/scatter flag to pattern vinfo.
9158 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
9160 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9161 behavior of vec_packsu (vector unsigned long long, vector unsigned
9162 long long) to match behavior of vec_packs with same signature.
9164 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
9165 Thomas Schwinge <thomas@codesourcery.com>
9166 Cesar Philippidis <cesar@codesourcery.com>
9168 * gimplify.c (gimplify_scan_omp_clauses): Add support for
9169 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9170 (gimplify_adjust_omp_clauses): Likewise.
9171 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
9172 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
9173 (gimplify_omp_target_update): Update handling of acc update and
9175 * omp-low.c (install_var_field): Remove unused parameter
9176 base_pointers_restrict.
9177 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
9178 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
9180 (omp_target_base_pointers_restrict_p): Delete.
9181 (scan_omp_target): Update call to scan_sharing_clauses.
9182 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
9184 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
9185 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9186 (convert_local_omp_clauses): Likewise.
9187 * tree-pretty-print.c (dump_omp_clause): Likewise.
9188 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
9190 (omp_clause_code_name): Likewise.
9192 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9195 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
9198 PR tree-optimization/86231
9199 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
9200 anti-range don't overwrite *vr0min before using it to compute *vr0max.
9202 2018-06-20 Tom de Vries <tdevries@suse.de>
9204 PR tree-optimization/86097
9205 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
9206 iv type if signedness of iv type is not the same as that of *nit.
9208 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9210 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
9211 EDGE_EH edges, verify they are all EDGE_EH.
9213 2018-06-20 Maya Rashish <coypu@sdf.org>
9215 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
9217 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9219 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
9220 * config/aarch64/aarch64.c (xgene1_tunings): Add
9221 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
9222 (aarch64_mode_valid_for_sched_fusion_p):
9223 Allow 16-byte modes.
9224 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
9225 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
9227 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
9229 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
9230 * config/aarch64/iterators.md (VQ2): New mode iterator.
9232 2018-06-20 Martin Liska <mliska@suse.cz>
9234 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9235 Change default ratio from 10 to 8.
9237 2018-06-20 Martin Liska <mliska@suse.cz>
9239 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9241 (bit_test_cluster::find_bit_tests): Likewise.
9242 (switch_decision_tree::analyze_switch_statement): Find clusters.
9243 * tree-switch-conversion.h (struct jump_table_cluster): Document
9246 2018-06-20 Martin Liska <mliska@suse.cz>
9248 * tree-switch-conversion.c (switch_conversion::collect):
9249 Record m_uniq property.
9250 (switch_conversion::expand): Bail out for special conditions.
9251 (group_cluster::~group_cluster): New.
9252 (group_cluster::group_cluster): Likewise.
9253 (group_cluster::dump): Likewise.
9254 (jump_table_cluster::emit): New.
9255 (switch_decision_tree::fix_phi_operands_for_edges): New.
9256 (struct case_node): Remove struct.
9257 (jump_table_cluster::can_be_handled): New.
9258 (case_values_threshold): Moved to header.
9259 (reset_out_edges_aux): Likewise.
9260 (jump_table_cluster::is_beneficial): New.
9261 (bit_test_cluster::can_be_handled): Likewise.
9262 (add_case_node): Remove.
9263 (bit_test_cluster::is_beneficial): New.
9264 (case_bit_test::cmp): New.
9265 (bit_test_cluster::emit): New.
9266 (expand_switch_as_decision_tree_p): Remove.
9267 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
9268 (fix_phi_operands_for_edge): Likewise.
9269 (switch_decision_tree::analyze_switch_statement): New.
9270 (compute_cases_per_edge): Move ...
9271 (switch_decision_tree::compute_cases_per_edge): ... here.
9272 (try_switch_expansion): Likewise.
9273 (switch_decision_tree::try_switch_expansion): Likewise.
9274 (record_phi_operand_mapping): Likewise.
9275 (switch_decision_tree::record_phi_operand_mapping): Likewise.
9276 (emit_case_decision_tree): Likewise.
9277 (switch_decision_tree::emit): Likewise.
9278 (balance_case_nodes): Likewise.
9279 (switch_decision_tree::balance_case_nodes): Likewise.
9280 (dump_case_nodes): Likewise.
9281 (switch_decision_tree::dump_case_nodes): Likewise.
9282 (emit_jump): Likewise.
9283 (switch_decision_tree::emit_jump): Likewise.
9284 (emit_cmp_and_jump_insns): Likewise.
9285 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
9286 (emit_case_nodes): Likewise.
9287 (switch_decision_tree::emit_case_nodes): Likewise.
9288 (conditional_probability): Remove.
9289 * tree-switch-conversion.h (enum cluster_type): New.
9291 (struct cluster): Likewise.
9292 (cluster::cluster): Likewise.
9293 (struct simple_cluster): Likewise.
9294 (simple_cluster::simple_cluster): Likewise.
9295 (struct group_cluster): Likewise.
9296 (struct jump_table_cluster): Likewise.
9297 (struct bit_test_cluster): Likewise.
9298 (struct min_cluster_item): Likewise.
9299 (struct case_tree_node): Likewise.
9300 (case_tree_node::case_tree_node): Likewise.
9301 (jump_table_cluster::case_values_threshold): Likewise.
9302 (struct case_bit_test): Likewise.
9303 (struct switch_decision_tree): Likewise.
9304 (struct switch_conversion): Likewise.
9305 (switch_decision_tree::reset_out_edges_aux): Likewise.
9307 2018-06-20 Martin Liska <mliska@suse.cz>
9309 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
9310 (hoist_edge_and_branch_if_true): Likewise.
9311 (expand_switch_using_bit_tests_p): Likewise.
9312 (struct case_bit_test): Likewise.
9313 (case_bit_test_cmp): Likewise.
9314 (emit_case_bit_tests): Likewise.
9315 (switch_conversion::switch_conversion): New class.
9316 (struct switch_conv_info): Remove old struct.
9317 (collect_switch_conv_info): More to ...
9318 (switch_conversion::collect): ... this.
9319 (check_range): Likewise.
9320 (switch_conversion::check_range): Likewise.
9321 (check_all_empty_except_final): Likewise.
9322 (switch_conversion::check_all_empty_except_final): Likewise.
9323 (check_final_bb): Likewise.
9324 (switch_conversion::check_final_bb): Likewise.
9325 (create_temp_arrays): Likewise.
9326 (switch_conversion::create_temp_arrays): Likewise.
9327 (free_temp_arrays): Likewise.
9328 (gather_default_values): Likewise.
9329 (switch_conversion::gather_default_values): Likewise.
9330 (build_constructors): Likewise.
9331 (switch_conversion::build_constructors): Likewise.
9332 (constructor_contains_same_values_p): Likewise.
9333 (switch_conversion::contains_same_values_p): Likewise.
9334 (array_value_type): Likewise.
9335 (switch_conversion::array_value_type): Likewise.
9336 (build_one_array): Likewise.
9337 (switch_conversion::build_one_array): Likewise.
9338 (build_arrays): Likewise.
9339 (switch_conversion::build_arrays): Likewise.
9340 (gen_def_assigns): Likewise.
9341 (switch_conversion::gen_def_assigns): Likewise.
9342 (prune_bbs): Likewise.
9343 (switch_conversion::prune_bbs): Likewise.
9344 (fix_phi_nodes): Likewise.
9345 (switch_conversion::fix_phi_nodes): Likewise.
9346 (gen_inbound_check): Likewise.
9347 (switch_conversion::gen_inbound_check): Likewise.
9348 (process_switch): Use the newly created class.
9349 (switch_conversion::expand): New.
9350 (switch_conversion::~switch_conversion): New.
9351 * tree-switch-conversion.h: New file.
9353 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9355 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
9356 tree-vect-patterns.c.
9357 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
9358 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
9359 (vect_recog_sad_pattern): Likewise.
9360 (vect_recog_widen_sum_pattern): Likewise.
9361 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
9362 (vect_recog_widen_shift_pattern): Remove the type_in argument.
9363 (vect_recog_rotate_pattern): Likewise.
9364 (vect_recog_mult_pattern): Likewise.
9365 (vect_recog_vector_vector_shift_pattern): Likewise.
9366 (vect_recog_divmod_pattern): Likewise.
9367 (vect_recog_mixed_size_cond_pattern): Likewise.
9368 (vect_recog_bool_pattern): Likewise.
9369 (vect_recog_mask_conversion_pattern): Likewise.
9370 (vect_try_gather_scatter_pattern): Likewise.
9371 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
9372 (vect_recog_over_widening_pattern): Likewise.
9373 (vect_recog_gather_scatter_pattern): Likewise.
9374 (vect_recog_func_ptr): Move from tree-vectorizer.h
9375 (vect_vect_recog_func_ptrs): Move further down the file.
9376 (vect_recog_func): Likewise. Remove the third argument.
9377 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
9378 (vect_pattern_recog_1): Expect the pattern function to do any
9379 necessary target tests. Also expect it to provide a vector type.
9380 Remove the type_in handling.
9382 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9384 * tree-vect-patterns.c (vect_pattern_detected): New function.
9385 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
9386 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
9387 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
9388 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
9389 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9390 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9391 (vect_recog_mask_conversion_pattern)
9392 (vect_try_gather_scatter_pattern): Likewise.
9394 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9396 * tree-vect-patterns.c (vect_get_internal_def): New function.
9397 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9398 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
9399 (search_type_for_mask_1): Use it.
9401 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9403 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9404 redundant WIDEN_SUM_EXPR handling.
9405 (vect_recog_sad_pattern): Likewise.
9407 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9409 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9410 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
9411 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
9412 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
9413 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
9415 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9417 * tree-vect-stmts.c (vectorizable_call): Make sure that we
9418 use the stmt_vec_info of the original bb statement for the
9419 new zero assignment, even if the call is part of a pattern.
9421 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9423 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
9424 that the sequence is attached to the original statement rather
9425 than the pattern statement.
9426 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
9427 PATTERN_DEF_SEQ from the original statement rather than
9428 the main pattern statement.
9429 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
9430 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
9431 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
9433 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9435 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
9436 definition statements before the early exit for statements that aren't
9438 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
9440 (vect_transform_loop): ...here. Process pattern definition
9441 statements without first checking whether the main pattern
9442 statement is live or relevant.
9444 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9446 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9447 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9449 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
9451 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9452 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9453 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9454 (expand_block_compare): Change select_block_compare_mode call.
9455 (expand_strncmp_align_check): Use new functions, fix comment.
9456 (emit_final_str_compare_gpr): New function.
9457 (expand_strn_compare): Refactor and clean up code.
9458 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9460 2018-06-19 Tony Reix <tony.reix@atos.com>
9461 Damien Bergamini <damien.bergamini@atos.com>
9462 David Edelsohn <dje.gcc@gmail.com>
9464 * collect2.c (static_obj): New variable.
9465 (static_libs): New variable.
9466 (is_in_list): Uncomment declaration.
9467 (main): Track AIX libraries linked statically.
9468 (is_in_list): Uncomment definition.
9469 (scan_prog_file): Don't add AIX shared libraries initializer
9470 to constructor list if linking statically.
9472 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
9474 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9476 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9478 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9480 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9483 2018-06-19 Martin Liska <mliska@suse.cz>
9485 * config/i386/i386.c (ix86_can_inline_p): Do not use
9486 ipa_fn_summaries::get_create.
9487 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9489 (devirtualization_time_bonus): Likewise.
9490 (ipcp_propagate_stage): Likewise.
9491 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9492 (edge_set_predicate): Likewise.
9493 (evaluate_conditions_for_known_args): Likewise.
9494 (evaluate_properties_for_edge): Likewise.
9495 (ipa_call_summary::reset): Tranform to ...
9496 (ipa_call_summary::~ipa_call_summary): ... this.
9497 (ipa_fn_summary::reset): Transform to ...
9498 (ipa_fn_summary::~ipa_fn_summary): ... this.
9499 (ipa_fn_summary_t::remove): Rename to ...
9500 (ipa_fn_summary_t::remove_callees): ... this.
9501 (ipa_fn_summary_t::duplicate): Use placement new
9502 instead of memory copy.
9503 (ipa_call_summary_t::duplicate): Likewise.
9504 (ipa_call_summary_t::remove): Remove.
9505 (dump_ipa_call_summary): Change get_create to get.
9506 (ipa_dump_fn_summary): Dump only when summary exists.
9507 (analyze_function_body): Use symbol_summary::get instead
9509 (compute_fn_summary): Likewise.
9510 (estimate_edge_devirt_benefit): Likewise.
9511 (estimate_edge_size_and_time): Likewise.
9512 (inline_update_callee_summaries): Likewise.
9513 (remap_edge_change_prob): Likewise.
9514 (remap_edge_summaries): Likewise.
9515 (ipa_merge_fn_summary_after_inlining): Likewise.
9516 (write_ipa_call_summary): Likewise.
9517 (ipa_fn_summary_write): Likewise.
9518 (ipa_free_fn_summary): Likewise.
9519 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9520 (struct ipa_call_summary): Likewise.
9521 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9523 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9524 (estimate_size_after_inlining): Likewise.
9525 (estimate_growth): Likewise.
9526 (growth_likely_positive): Likewise.
9527 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9528 (inline_call): Likewise.
9529 * ipa-inline.c (caller_growth_limits): Likewise.
9530 (can_inline_edge_p): Likewise.
9531 (can_inline_edge_by_limits_p): Likewise.
9532 (compute_uninlined_call_time): Likewise.
9533 (compute_inlined_call_time): Likewise.
9534 (want_inline_small_function_p): Likewise.
9535 (edge_badness): Likewise.
9536 (update_caller_keys): Likewise.
9537 (update_callee_keys): Likewise.
9538 (inline_small_functions): Likewise.
9539 (inline_to_all_callers_1): Likewise.
9540 (dump_overall_stats): Likewise.
9541 (early_inline_small_functions): Likewise.
9542 (early_inliner): Likewise.
9543 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9544 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9545 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9546 * ipa-split.c (execute_split_functions): Likewise.
9547 * symbol-summary.h: Likewise.
9548 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9550 2018-06-19 Richard Biener <rguenther@suse.de>
9552 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9553 (vectorize_loops): ... here. Fix dbgcnt handling.
9554 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9556 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9559 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9560 ieee128 argument takes up only one (vector) register, not two (floating
9563 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9565 * gimplify.c (gimplify_init_constructor): Really never clear for an
9566 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9568 2018-06-19 Richard Biener <rguenther@suse.de>
9570 PR tree-optimization/86179
9571 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9572 after failed recognition.
9574 2018-06-18 Martin Sebor <msebor@redhat.com>
9577 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9578 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9579 Handle integer subtraction.
9580 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9581 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9583 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9585 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9586 param from rtx to rtx_insn *.
9587 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9589 (frv_ifcvt_modify_insn): Likwise.
9590 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9591 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9592 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9593 as_a <rtx_insn *> cast to local "unprotected_region" once
9594 it's been established that it's not NULL or pc_rtx.
9595 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9596 param "sethi" from rtx to rtx_insn *.
9597 (nds32_group_float_insns): Likewise for param "insn".
9598 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9600 (vax_output_int_subtract): Likewise.
9601 * config/vax/vax.c (vax_output_int_add): Likewise for param
9603 (vax_output_int_subtract): Likewise.
9604 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9605 (emit_pattern_after): Likewise for param "after".
9606 (emit_insn_after): Likewise.
9607 (emit_jump_insn_after): Likewise.
9608 (emit_call_insn_after): Likewise.
9609 (emit_debug_insn_after): Likewise.
9610 (emit_pattern_before): Likewise for param "before".
9611 (emit_insn_before): Likewise.
9612 (emit_jump_insn_before): Likewise.
9613 * final.c (get_insn_template): Likewise for param "insn", removing
9615 * output.h (get_insn_template): Likewise for 2nd param.
9616 * rtl.h (emit_insn_before): Likewise.
9617 (emit_jump_insn_before): Likewise.
9618 (emit_debug_insn_before_noloc): Likewise.
9619 (emit_insn_after): Likewise.
9620 (emit_jump_insn_after): Likewise.
9621 (emit_call_insn_after): Likewise.
9622 (emit_debug_insn_after): Likewise.
9623 (set_insn_deleted): Likewise for param.
9625 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9628 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9629 floating point modes, so that IFmode is numerically greater than
9630 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9631 to declare the ordering. This prevents IFmode from being
9632 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9633 machine. Include rs6000-modes.h to share the fractional values
9634 between genmodes* and the rest of the compiler.
9638 * config/rs6000/rs6000-modes.h: New file.
9639 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9640 meaning of rs6000_long_double_size so that 126..128 selects an
9641 appropriate 128-bit floating point type.
9642 (rs6000_option_override_internal): Likewise.
9643 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9644 (TARGET_LONG_DOUBLE_128): Change the meaning of
9645 rs6000_long_double_size so that 126..128 selects an appropriate
9646 128-bit floating point type.
9647 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9648 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9649 source and destination to match the standard usage.
9650 (truncifkf2): Likewise.
9651 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9652 ISA 2.07 to use an explicit clobber, instead of passing in a
9654 (copysign<mode>3_soft): Likewise.
9656 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9658 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9659 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9660 (vect_slp_analyze_instance_dependence): Likewise.
9661 (vect_enhance_data_refs_alignment): Likewise.
9662 (vect_analyze_data_refs_alignment): Likewise.
9663 (vect_slp_analyze_and_verify_instance_alignment
9664 (vect_analyze_data_ref_accesses): Likewise.
9665 (vect_prune_runtime_alias_test_list): Likewise.
9666 (vect_analyze_data_refs): Likewise.
9667 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9668 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9669 (vect_analyze_scalar_cycles_1): Likewise.
9670 (vect_get_loop_niters): Likewise.
9671 (vect_analyze_loop_form_1): Likewise.
9672 (vect_update_vf_for_slp): Likewise.
9673 (vect_analyze_loop_operations): Likewise.
9674 (vect_analyze_loop): Likewise.
9675 (vectorizable_induction): Likewise.
9676 (vect_transform_loop): Likewise.
9677 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9678 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9679 (vect_make_slp_decision): Likewise.
9680 (vect_detect_hybrid_slp): Likewise.
9681 (vect_slp_analyze_operations): Likewise.
9682 (vect_slp_bb): Likewise.
9683 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9684 (vectorizable_bswap): Likewise.
9685 (vectorizable_call): Likewise.
9686 (vectorizable_simd_clone_call): Likewise.
9687 (vectorizable_conversion): Likewise.
9688 (vectorizable_assignment): Likewise.
9689 (vectorizable_shift): Likewise.
9690 (vectorizable_operation): Likewise.
9691 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9693 2018-06-18 Martin Sebor <msebor@redhat.com>
9695 PR tree-optimization/81384
9696 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9697 * builtins.c (expand_builtin_strnlen): New function.
9698 (expand_builtin): Call it.
9699 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9700 * builtins.def (BUILT_IN_STRNLEN): New.
9701 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9702 Warn for bounds in excess of maximum object size.
9703 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9704 single-value ranges. Handle strnlen.
9705 (handle_builtin_strlen): Handle strnlen.
9706 (strlen_check_and_optimize_stmt): Same.
9707 * doc/extend.texi (Other Builtins): Document strnlen.
9709 2018-06-18 Maya Rashish <coypu@sdf.org>
9711 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9712 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9713 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9715 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9717 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9719 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9721 * tree.c (escaped_string::escape): Replace cast to char * by
9722 const_cast<char *> (unescaped).
9724 2018-06-18 Nick Clifton <nickc@redhat.com>
9727 * tree.c (escaped_string): New class. Converts an unescaped
9728 string into its escaped equivalent.
9729 (warn_deprecated_use): Use the new class to convert the
9730 deprecation message, if present.
9731 (test_escaped_strings): New self test.
9732 (test_c_tests): Add test_escaped_strings.
9733 * doc/extend.texi (deprecated): Add a note that the
9734 deprecation message is affected by the -fmessage-length
9735 option, and that control characters will be escaped.
9736 (#pragma GCC error): Document this pragma.
9737 (#pragma GCC warning): Likewise.
9738 * doc/invoke.texi (-fmessage-length): Document this option's
9739 effect on the #warning and #error preprocessor directives and
9740 the deprecated attribute.
9742 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
9744 * tree.c (decl_value_expr_lookup): Revert latest change.
9745 (decl_value_expr_insert): Likewise.
9747 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
9749 * gimplify.c (nonlocal_vlas): Delete.
9750 (nonlocal_vla_vars): Likewise.
9751 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9753 (gimplify_body): Do not create and destroy nonlocal_vlas.
9754 * tree-nested.c: Include diagnostic.h.
9755 (use_pointer_in_frame): Tweak.
9756 (lookup_field_for_decl): Add assertion and declare the transformation.
9757 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9758 internal error when the reference is in a wrong context. Do not
9759 create a debug decl by default.
9760 (note_nonlocal_block_vlas): Delete.
9761 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9762 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
9763 create a debug decl by default.
9764 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9765 call is in a wrong context.
9766 (fixup_vla_decls): New function.
9767 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
9768 debug variables were created.
9769 * tree.c (decl_value_expr_lookup): Add checking assertion.
9770 (decl_value_expr_insert): Likewise.
9772 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9775 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9776 * tree-scalar-evolution.c (interpret_expr): Likewise.
9777 (expression_expensive_p): Likewise.
9778 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9779 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9780 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9781 (ssa_defined_by_minus_one_stmt_p): New.
9783 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9786 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9787 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9788 * dojump.c (do_jump): Likewise.
9789 * expr.c (expand_expr_real_2): Check operand type's sign.
9790 * fold-const.c (const_unop): Handle ABSU_EXPR.
9791 (fold_abs_const): Likewise.
9792 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9793 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9794 (strip_sign_op_1): Likesise.
9795 * match.pd: Add new pattern to generate ABSU_EXPR.
9796 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9797 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9798 * tree-eh.c (operation_could_trap_helper_p): Likewise.
9799 * tree-inline.c (estimate_operator_cost): Likewise.
9800 * tree-pretty-print.c (dump_generic_node): Likewise.
9801 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9802 * tree.def (ABSU_EXPR): New.
9804 2018-06-16 Jakub Jelinek <jakub@redhat.com>
9807 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9808 documented as preserved for backward compatibility only.
9809 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9811 PR rtl-optimization/86108
9812 * bb-reorder.c (create_forwarder_block): Renamed to ...
9813 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9814 jump from new landing pad to the second part.
9815 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9818 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9821 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9822 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9823 Only call store_expr for halves if the mode is the same.
9826 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9827 Fix up comment formatting.
9829 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9831 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9832 (typed_splay_tree::closure,
9833 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9834 (typed_splay_tree::typed_splay_tree,
9835 typed_splay_tree::operator =): Declared private.
9836 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9837 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9838 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9839 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9840 typed_splay_tree::splay_tree_splay,
9841 typed_splay_tree::splay_tree_foreach_helper,
9842 typed_splay_tree::splay_tree_insert,
9843 typed_splay_tree::splay_tree_remove,
9844 typed_splay_tree::splay_tree_lookup,
9845 typed_splay_tree::splay_tree_predecessor,
9846 typed_splay_tree::splay_tree_successor,
9847 typed_splay_tree::splay_tree_min,
9848 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9849 (typed_splay_tree::root, typed_splay_tree::comp,
9850 typed_splay_tree::delete_key,
9851 typed_splay_tree::delete_value): New data members.
9852 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9853 typed_splay_tree::remove.
9855 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9857 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9858 -mginv and -mno-ginv to the assembler.
9859 * config/mips/mips.opt (-mcrc): New option.
9861 * doc/invoke.text (-mcrc): Document.
9864 2018-06-15 Nick Clifton <nickc@redhat.com>
9867 * tree.c (escaped_string): New class. Converts an unescaped
9868 string into its escaped equivalent.
9869 (warn_deprecated_use): Use the new class to convert the
9870 deprecation message, if present.
9871 (test_escaped_strings): New self test.
9872 (test_c_tests): Add test_escaped_strings.
9873 * doc/extend.texi (deprecated): Add a note that the
9874 deprecation message is affected by the -fmessage-length
9875 option, and that control characters will be escaped.
9876 (#pragma GCC error): Document this pragma.
9877 (#pragma GCC warning): Likewise.
9878 * doc/invoke.texi (-fmessage-length): Document this option's
9879 effect on the #warning and #error preprocessor directives and
9880 the deprecated attribute.
9882 2018-06-15 Richard Biener <rguenther@suse.de>
9884 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9885 here, also noting vector size used.
9886 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9887 size used in MSG_OPTIMIZED_LOCATIONS dump.
9888 (pass_slp_vectorize::execute): Adjust.
9890 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9893 * config/arc/arc.c (arc_return_address_register): Fix
9896 2018-06-15 Richard Biener <rguenther@suse.de>
9899 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9900 leave useless conversion stripping to force_gimple_operand_gsi.
9901 (gimplify_build2): Likewise.
9902 (gimplify_build1): Likewise.
9904 2018-06-15 Richard Biener <rguenther@suse.de>
9907 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9908 before adjusting their block.
9910 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9912 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9913 multilibs for *-*-rtems*.
9914 * config/riscv/t-rtems: New file.
9916 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9919 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9920 unsigned_type_for returns NULL.
9923 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9924 subregs of multi-word pseudos unless the float mode has word size.
9926 2018-06-14 Richard Biener <rguenther@suse.de>
9929 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9930 duplicate and harmful type_hash_canon.
9931 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9933 2018-06-14 Richard Biener <rguenther@suse.de>
9936 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9939 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9941 * config/rtems.h (STDINT_LONG32): Define.
9943 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9944 Prachi Godbole <prachi.godbole@imgtec.com>
9946 * config/mips/mips-cpus.def: Define P6600.
9947 * config/mips/mips-tables.opt: Regenerate.
9948 * config/mips/mips.c (mips_ucbranch_type): New enum.
9949 (mips_rtx_cost_data): Add support for P6600.
9950 (mips_issue_rate): Likewise.
9951 (mips_multipass_dfa_lookahead): Likewise.
9952 (mips_avoid_hazard): Likewise.
9953 (mips_reorg_process_insns): Likewise.
9954 (mips_classify_branch_p6600): New function.
9955 * config/mips/mips.h (TUNE_P6600): New define.
9956 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9957 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9958 * config/mips/mips.md: Include p6600.md.
9959 (processor): Add p6600.
9960 * config/mips/p6600.md: New file.
9961 * doc/invoke.texi: Add p6600 to supported architectures.
9963 2018-06-13 Martin Sebor <msebor@redhat.com>
9965 PR tree-optimization/86114
9966 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9968 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9970 2018-06-13 Richard Biener <rguenther@suse.de>
9972 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9973 Properly set vector type of the intermediate stmt.
9974 * tree-vect-stmts.c (vectorizable_operation): The destination
9975 var always has vectype_out type.
9977 2018-06-13 Jeff Law <law@redhat.com>
9979 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9980 integer 0 for argument to print_rtl_with_bb.
9981 (rl78_reorg): Likewise.
9983 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9985 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9986 from rtx to rtx_insn *.
9987 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9989 (add_sched_insns_for_speculation): Likewise for local "target",
9990 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9991 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9992 from rtx_insn ** to rtx_code_label **.
9993 (reorg_emit_nops): Likewise.
9994 (c6x_reorg): Likewise for local "call_labels".
9995 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9997 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9998 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9999 the loops over LABEL_REFS.
10000 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
10002 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10003 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
10004 (split_branches): Strengthen local "olabel" from rtx to
10005 rtx_insn *, adding a safe_as_a cast.
10006 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
10008 (add_insn_after): Likewise for first two params.
10009 (add_insn_before): Likewise.
10010 (remove_insn): Likewise for param.
10011 (emit_pattern_before_noloc): Likewise for second and third params.
10012 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
10013 (emit_call_insn_before_noloc): Likewise.
10014 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
10016 (emit_barrier_before): Likewise.
10017 (emit_label_before): Strengthen "label" param from "rtx" to
10018 "rtx_code_label *". Strengthen "before" param from "rtx" to
10020 (emit_insn_after_1): Strengthen "after" param from "rtx" to
10022 (emit_pattern_after_noloc): Likewise.
10023 (emit_insn_after_noloc): Likewise.
10024 (emit_jump_insn_after_noloc): Likewise.
10025 (emit_call_insn_after_noloc): Likewise.
10026 (emit_debug_insn_after_noloc): Likewise.
10027 (emit_barrier_after): Likewise.
10028 (emit_label_after): Likewise for both params.
10029 (emit_pattern_after_setloc): Likewise for "after" param. Convert
10030 "loc" param from "int" to "location_t".
10031 (emit_insn_after_setloc): Likewise.
10032 (emit_jump_insn_after_setloc): Likewise.
10033 (emit_call_insn_after_setloc): Likewise.
10034 (emit_debug_insn_after_setloc): Likewise.
10035 (emit_pattern_before_setloc): Likewise for "before" param. Convert
10036 "loc" param from "int" to "location_t".
10037 (emit_pattern_before): Convert NULL_RTX to NULL.
10038 (emit_insn_before_setloc): Convert "loc" param from "int" to
10040 (emit_jump_insn_before_setloc): Likewise.
10041 (emit_call_insn_before_setloc): Likewise.
10042 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
10043 rtx_insn *. Convert "loc" param from "int" to "location_t".
10044 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
10045 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
10046 Convert 3rd param from "int" to "location_t".
10047 (emit_barrier_before, emit_barrier_after, next_real_insn):
10048 Strengthen param from rtx to rtx_insn *.
10049 (emit_label_before): Strengthen 1st param from "rtx" to
10050 "rtx_code_label *". Strengthen 2nd param from "rtx" to
10052 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
10053 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
10054 Strengthen 2nd param from "rtx" to "rtx_insn *".
10055 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
10056 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
10057 Likewise. Convert 3rd param from "int" to "location_t".
10058 (emit_label_after): Strengthen 1st param from "rtx" to
10059 "rtx_code_label *".
10060 (next_real_insn, remove_insn): Strengthen param from "rtx" to
10062 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
10063 from "rtx" to "rtx_insn *".
10065 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
10067 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
10068 bodies streamed in with -Q.
10069 * dumpfile.c (dump_files): Add lto-stream-out dump file.
10070 * dumpfile.h (tree_dump_index): Add lto_stream_out.
10071 * gimple-streamer-out.c: Include gimple-pretty-print.h
10072 (output_bb): Dump stmts streamed.
10073 * lto-section-out.c: Include print-tree.h
10074 (lto_begin_section): Dump sections created.
10075 (lto_output_decl_index): Dump decl encoded.
10076 * lto-streamer-out.c: Include print-tree.h
10077 (create_output_block): Dump output block created.
10078 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
10079 (output_function): Dump function output.
10080 (output_constructor): Dump constructor streamed.
10081 (write_global_stream): Output indexes encoded.
10082 (produce_asm_for_decls): Dump streams encoded.
10083 * lto-streamer.c (streamer_dump_file): New global var.
10084 * lto-streamer.h (streamer_dump_file): Declare.
10085 * passes.c (ipa_write_summaries): Initialize streamer dump.
10086 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
10089 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
10092 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
10093 offsets for register save directives. Emit a second batch of save
10094 directives, if need be, when the function accesses prior frames.
10096 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10098 * config/arc/fpu.md (fmasf4): Force operand to register.
10099 (fnmasf4): Likewise.
10101 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10103 * config/arc/arc-protos.h (arc_pad_return): Remove.
10104 * config/arc/arc.c (machine_function): Remove force_short_suffix
10106 (arc_print_operand): Adjust printing of '&'.
10107 (arc_verify_short): Remove conditional printing of short suffix.
10108 (arc_final_prescan_insn): Remove reference to size_reason.
10109 (pad_return): New function.
10110 (arc_reorg): Call pad_return.
10111 (arc_pad_return): Remove.
10112 (arc_init_machine_status): Remove reference to force_short_suffix.
10113 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
10114 (attr length): When attribute iscompact is true force to 2
10115 regardless; in the case of maybe check if we want to force the
10116 instruction to have 4 bytes length.
10117 (nopv): Change it to generate 4 byte long nop as well.
10118 (blockage): New pattern.
10119 (simple_return): Remove call to arc_pad_return.
10120 (p_return_i): Likewise.
10122 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10124 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
10126 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10128 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
10131 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10133 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
10134 for ARC700 and ARCv2.
10136 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
10139 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
10140 operands[2] instead of operands[1].
10143 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10145 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
10146 case, check whether the outer register overlaps an unallocatable
10147 register, not just whether it fits the required class.
10149 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10151 * poly-int.h (can_div_trunc_p): Add new overload in which all values
10153 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
10154 (memrefs_conflict_p): Likewise.
10155 (init_alias_analysis): Likewise.
10156 * cfgexpand.c (expand_debug_expr): Likewise.
10157 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10158 * cse.c (fold_rtx): Likewise.
10159 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
10160 * expr.c (emit_block_move_hints): Likewise.
10161 (clear_storage_hints, push_block, emit_push_insn): Likewise.
10162 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
10163 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
10164 (emit_group_store): Likewise.
10165 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
10166 to read the PRE/POST_MODIFY increment.
10167 * calls.c (store_one_arg): Use strip_offset.
10168 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
10170 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
10172 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
10173 (simplify_binary_operation_1): Extend CONST_INT handling to
10175 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
10176 than a HOST_WIDE_INT.
10177 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
10179 (adjust_mems, add_stores): Update accodingly.
10180 (vt_canonicalize_addr): Track polynomial offsets.
10181 (emit_note_insn_var_location): Likewise.
10182 (vt_add_function_parameter): Likewise.
10183 (vt_initialize): Likewise.
10185 2018-06-12 Jeff Law <law@redhat.com>
10187 * config.gcc (alpha*-*-freebsd*): Remove.
10188 * config/alpha/freebsd.h: Remove.
10190 2018-06-12 David Malcolm <dmalcolm@redhat.com>
10193 * spellcheck-tree.c (levenshtein_distance): Rename to...
10194 (get_edit_distance): ...this, and update for underlying renaming.
10195 * spellcheck-tree.h (levenshtein_distance): Rename to...
10196 (get_edit_distance): ...this.
10197 * spellcheck.c (levenshtein_distance): Rename to...
10198 (get_edit_distance): ...this. Convert from Levenshtein distance
10199 to Damerau-Levenshtein distance by supporting transpositions of
10200 adjacent characters. Rename "v1" to "v_next" and "v0" to
10202 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
10203 (selftest::test_edit_distance_unit_test_oneway): ...this, and
10204 update for underlying renaming.
10205 (selftest::levenshtein_distance_unit_test): Rename to...
10206 (selftest::test_get_edit_distance_unit): ...this, and update for
10207 underlying renaming.
10208 (selftest::test_find_closest_string): Add example from PR 69968
10209 where transposition helps
10210 (selftest::test_metric_conditions): Update for renaming.
10211 (selftest::test_metric_conditions): Likewise.
10212 (selftest::spellcheck_c_tests): Likewise.
10213 * spellcheck.h (levenshtein_distance): Rename both overloads to...
10214 (get_edit_distance): ...this.
10215 (best_match::consider): Update for renaming.
10217 2018-06-12 Martin Sebor <msebor@redhat.com>
10219 PR tree-optimization/85259
10220 * builtins.c (compute_objsize): Handle constant offsets.
10221 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
10222 true iff a warning has been issued.
10223 * gimple.h (gimple_nonartificial_location): New function.
10224 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
10225 gimple_nonartificial_location and handle -Wno-system-headers.
10226 (handle_builtin_stxncpy): Same.
10228 2018-06-12 Martin Sebor <msebor@redhat.com>
10231 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
10233 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
10235 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
10236 BUILTIN_VEC_XST entries for pointer to double and long long.
10238 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
10241 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10243 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10246 2018-06-12 Martin Liska <mliska@suse.cz>
10248 * doc/options.texi: Document IntegerRange.
10250 2018-06-12 Martin Liska <mliska@suse.cz>
10252 * config/i386/i386.opt: Make MPX-related options as Deprecated.
10253 * opt-functions.awk: Handle Deprecated flag.
10254 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
10256 (read_cmdline_option): Report warning for a deprecated option.
10257 * opts.h (struct cl_option): Add new field cl_deprecated.
10258 (CL_ERR_DEPRECATED): New.
10260 2018-06-12 Martin Liska <mliska@suse.cz>
10262 * doc/options.texi: Document Deprecated option flag.
10264 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10266 * config/arc/arc-arch.h (arc_extras): New enum.
10267 (arc_cpu_t):Add field extra.
10268 (arc_cpu_types): Consider the extras.
10269 * config/arc/arc-cpus.def: Add extras info.
10270 * config/arc/arc-opts.h (processor_type): Consider extra field.
10271 * config/arc/arc.c (arc_override_options): Handle extra field.
10273 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10275 * config/arc/arc-arch.h: Update ARC_OPTX macro.
10276 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
10278 * config/arc/arc.c (arc_init): Update pic warning.
10279 (irq_range): Update irq range parsing warnings.
10280 (arc_override_options): Update various warning messages.
10281 (arc_handle_aux_attribute): Likewise.
10283 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
10285 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
10287 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10289 * doc/sourcebuild.texi: Document usage of line number 0 in verify
10290 compiler messages directives.
10292 2018-06-12 Matthew Fortune <mfortune@gmail.com>
10294 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
10295 * config/mips/mips-tables.opt: Regenerate.
10296 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
10298 * doc/invoke.texi: Document -march=i6500.
10300 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
10302 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
10303 (i6400_gpmul): Add cpu_unit.
10304 (i6400_gpdiv): Likewise.
10305 (i6400_msa_add_d): Update reservations.
10306 (i6400_msa_int_add) Likewise.
10307 (i6400_msa_short_logic3) Likewise.
10308 (i6400_msa_short_logic2) Likewise.
10309 (i6400_msa_short_logic) Likewise.
10310 (i6400_msa_move) Likewise.
10311 (i6400_msa_cmp) Likewise.
10312 (i6400_msa_short_float2) Likewise.
10313 (i6400_msa_div_d) Likewise.
10314 (i6400_msa_long_logic1) Likewise.
10315 (i6400_msa_long_logic2) Likewise.
10316 (i6400_msa_mult) Likewise.
10317 (i6400_msa_long_float2) Likewise.
10318 (i6400_msa_long_float4) Likewise.
10319 (i6400_msa_long_float5) Likewise.
10320 (i6400_msa_long_float8) Likewise.
10321 (i6400_fpu_fadd): Include frint type.
10322 (i6400_fpu_store): New define_insn_reservation.
10323 (i6400_fpu_load): Likewise.
10324 (i6400_fpu_move): Likewise.
10325 (i6400_fpu_fcmp): Likewise.
10326 (i6400_fpu_fmadd): Likewise.
10327 (i6400_int_mult): Include imul3nc type and update reservation.
10328 (i6400_int_div): Include idiv3 type and update reservation.
10329 (i6400_int_load): Update to check type not move_type.
10330 (i6400_int_store): Likewise.
10331 (i6400_int_prefetch): Set zero latency.
10333 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
10335 * gcc.c: Document new %@{...} sequence.
10336 (LINK_COMMAND_SPEC): Use it for the -L switches.
10337 (cpp_unique_options): Use it for the -I switches.
10338 (at_file_argbuf): New global variable.
10339 (in_at_file): Likewise.
10340 (alloc_args): Create at_file_argbuf.
10341 (clear_args): Truncate at_file_argbuf.
10342 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
10343 (open_at_file): New function.
10344 (close_at_file): Likewise.
10345 (create_at_file): Delete.
10346 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
10349 (validate_switches_from_spec): Deal with %@{...} sequence.
10350 (validate_switches): Likewise.
10351 (driver::finalize): Call clear_args.
10353 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
10355 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
10357 2018-06-11 Martin Sebor <msebor@redhat.com>
10359 * doc/invoke.texi (-Wall): List -Wc++17-compat.
10360 (Wno-class-memaccess): Add @opindex.
10361 (Wno-templates, Wno-multiple-inheritance): Same.
10362 (Wno-virtual-inheritance, Wno-namespaces): Same.
10363 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
10364 (Wno-format-overflow, Wno-format-truncation): Same.
10365 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
10366 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
10367 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
10368 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
10369 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
10370 (Wno-misspelled-isr): Same.
10372 2018-06-11 Martin Sebor <msebor@redhat.com>
10374 * PR tree-optimization/86083
10375 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
10377 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
10379 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
10381 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
10384 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
10385 on the correct operand.
10386 (*movdi_internal64): Ditto.
10388 2018-06-11 Martin Liska <mliska@suse.cz>
10390 PR tree-optimization/86089
10391 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
10393 2018-06-11 Julia Koval <julia.koval@intel.com>
10395 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
10396 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
10397 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
10399 2018-06-11 Olivier Hainque <hainque@adacore.com>
10401 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
10402 for Ada with strict dwarf2.
10404 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
10407 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
10410 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
10412 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
10414 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
10416 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
10419 2018-06-08 Martin Liska <mliska@suse.cz>
10421 * tree-cfg.h (debug_function): Fix argument type to match
10424 2018-06-08 Martin Liska <mliska@suse.cz>
10426 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
10427 Remove usage of MPX-related (and removed) fields.
10428 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
10430 2018-06-08 David Malcolm <dmalcolm@redhat.com>
10432 * cfg.c (debug): Use TDF_NONE rather than 0.
10433 * cfghooks.c (debug): Likewise.
10434 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
10435 (struct dump_option_value_info): Convert to...
10436 (struct kv_pair): ...this template type.
10437 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
10439 (optinfo_verbosity_options): Likewise.
10440 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
10442 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10443 than int for "optgroup_flags" param.
10444 (dump_generic_expr_loc): Use dump_flags_t rather than int for
10446 (dump_dec): Likewise.
10447 (dump_finish): Use TDF_NONE rather than 0.
10448 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10449 rather than int for "optgroup_flags" param. Use TDF_NONE rather
10450 than 0. Update for change to option_ptr.
10451 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10452 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
10453 0. Update for changes to optinfo_verbosity_options and
10455 (opt_info_switch_p): Convert optgroup_flags from int to
10457 (dump_basic_block): Use dump_flags_t rather than int
10458 for "dump_kind" param.
10459 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10460 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10461 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10462 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10463 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10464 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10465 TDF_NONE): Convert from macros to...
10466 (enum dump_flag): ...this new enum.
10467 (dump_flags_t): Update to use enum.
10468 (operator|, operator&, operator~, operator|=, operator&=):
10469 Implement for dump_flags_t.
10470 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10471 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10472 Convert from macros to...
10473 (enum optgroup_flag): ...this new enum.
10474 (optgroup_flags_t): New typedef.
10475 (operator|, operator|=): Implement for optgroup_flags_t.
10476 (struct dump_file_info): Convert field "alt_flags" to
10477 dump_flags_t. Convert field "optgroup_flags" to
10479 (dump_basic_block): Use dump_flags_t rather than int for param.
10480 (dump_generic_expr_loc): Likewise.
10481 (dump_dec): Likewise.
10482 (dump_register): Convert param "optgroup_flags" to
10484 (opt_info_enable_passes): Likewise.
10485 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10487 * gimple-pretty-print.c (debug): Likewise.
10488 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10489 (merged_store_group::apply_stores): Likewise.
10490 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10491 * gimple.c (verify_gimple_pp): Likewise.
10492 * graphite-poly.c (print_pbb_body): Likewise.
10493 * passes.c (pass_manager::register_one_dump_file): Convert
10494 local "optgroup_flags" to optgroup_flags_t.
10495 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10497 (debug_body): Likewise.
10498 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10499 to optgroup_flags_t.
10500 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10502 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10503 (convert_mult_to_fma): Likewise.
10504 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10505 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10506 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10507 "dump_kind" to dump_flags_t.
10509 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10511 * config/rs6000/rs6000.c (min, max): Delete.
10513 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10515 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10516 -mabi=spe and -mabi=no-spe.
10518 2018-06-08 Martin Liska <mliska@suse.cz>
10520 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10521 where we expect an existing summary.
10523 2018-06-08 Martin Liska <mliska@suse.cz>
10525 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10526 * ipa-inline.h (estimate_edge_growth): Likewise.
10528 2018-06-08 Martin Liska <mliska@suse.cz>
10530 * cgraph.c (function_version_hasher::hash): Use
10531 cgraph_node::get_uid ().
10532 (function_version_hasher::equal):
10533 * cgraph.h (cgraph_node::get_uid): New method.
10534 * ipa-inline.c (update_caller_keys): Use
10535 cgraph_node::get_uid ().
10536 (update_callee_keys): Likewise.
10537 * ipa-utils.c (searchc): Likewise.
10538 (ipa_reduced_postorder): Likewise.
10539 * lto-cgraph.c (input_node): Likewise.
10540 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10541 * symbol-summary.h (symtab_insertion): Likewise.
10542 (symtab_removal): Likewise.
10543 (symtab_duplication): Likewise.
10544 * tree-pretty-print.c (dump_function_header): Likewise.
10545 * tree-sra.c (convert_callers_for_node): Likewise.
10547 2018-06-08 Martin Liska <mliska@suse.cz>
10549 * cgraph.c (symbol_table::create_edge): Always assign a new
10551 (symbol_table::free_edge): Do not recycle numbers.
10552 * cgraph.h (cgraph_edge::get): New method.
10553 * symbol-summary.h (symtab_removal): Use it.
10554 (symtab_duplication): Likewise.
10555 (call_summary::hashable_uid): Remove.
10557 2018-06-08 Martin Liska <mliska@suse.cz>
10559 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10560 (initialize_growth_caches): Remove.
10561 (free_growth_caches): Likewise.
10562 (do_estimate_edge_time): Use edge_growth_cache.
10563 (do_estimate_edge_size): Likewise.
10564 (do_estimate_edge_hints): Likewise.
10565 * ipa-inline.c (reset_edge_caches): Likewise.
10566 (recursive_inlining): Likewise.
10567 (inline_small_functions): Likewise.
10568 * ipa-inline.h (initialize_growth_caches): Remove.
10569 (estimate_edge_size): Likewise.
10570 (estimate_edge_time): Likewise.
10571 (estimate_edge_hints): Likewise.
10572 (reset_edge_growth_cache): Likewise.
10573 * symbol-summary.h (call_summary::remove): New method.
10575 2018-06-08 Martin Liska <mliska@suse.cz>
10577 * ipa-cp.c (class edge_clone_summary): New summary.
10578 (grow_edge_clone_vectors): Remove.
10579 (ipcp_edge_duplication_hook): Remove.
10580 (class edge_clone_summary_t): New call_summary class.
10581 (ipcp_edge_removal_hook): Remove.
10582 (edge_clone_summary_t::duplicate): New function.
10583 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10584 (create_specialized_node): Likewise.
10585 (ipcp_driver): Initialize edge_clone_summaries and do not
10588 2018-06-08 Martin Liska <mliska@suse.cz>
10590 * symbol-summary.h (get): New function.
10591 (call_summary::m_initialize_when_cloning): New class member.
10593 2018-06-08 Martin Liska <mliska@suse.cz>
10595 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10596 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10597 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10598 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10599 (remove_cgraph_node_from_order): Use the removed_nodes set.
10600 (do_per_function_toporder): Likwise.
10601 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10602 instead of summary_uid.
10603 (symtab_removal): Likewise.
10604 (symtab_duplication): Likewise.
10606 2018-06-08 Martin Liska <mliska@suse.cz>
10608 * ipa-cp.c (ipcp_store_bits_results): Use
10609 ipcp_transformation_sum.
10610 (ipcp_store_vr_results): Likewise.
10611 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10613 (ipcp_transformation_initialize): ... this.
10614 (ipa_set_node_agg_value_chain):
10615 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10616 (write_ipcp_transformation_info): Likewise.
10617 (read_ipcp_transformation_info): Likewise.
10618 (ipcp_update_bits): Likewise.
10619 (ipcp_update_vr): Likewise.
10620 (ipcp_transform_function): Likewise.
10621 * ipa-prop.h: Rename ipcp_transformation_summary to
10622 ipcp_transformation.
10623 (class ipcp_transformation_t): New function summary.
10624 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10625 (ipa_get_agg_replacements_for_node): Likewise.
10627 2018-06-08 Martin Liska <mliska@suse.cz>
10629 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10631 (class funct_state_summary_t): New function_summary class.
10632 (has_function_state): Remove.
10633 (get_function_state): Likewise.
10634 (set_function_state): Likewise.
10635 (add_new_function): Likewise.
10636 (funct_state_summary_t::insert): New function.
10637 (duplicate_node_data): Remove.
10638 (remove_node_data): Remove.
10639 (funct_state_summary_t::duplicate): New function.
10640 (register_hooks): Create new funct_state_summaries.
10641 (pure_const_generate_summary): Use it.
10642 (pure_const_write_summary): Likewise.
10643 (pure_const_read_summary): Likewise.
10644 (propagate_pure_const): Likewise.
10645 (propagate_nothrow): Likewise.
10646 (dump_malloc_lattice): Likewise.
10647 (propagate_malloc): Likewise.
10648 (execute): Do not register hooks, just remove summary
10650 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10653 2018-06-08 Martin Liska <mliska@suse.cz>
10655 * ipa-reference.c (remove_node_data): Remove.
10656 (duplicate_node_data): Likewise.
10657 (class ipa_ref_var_info_summary_t): New class.
10658 (class ipa_ref_opt_summary_t): Likewise.
10659 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10660 (get_reference_optimization_summary): Use
10661 ipa_ref_opt_sum_summaries.
10662 (set_reference_vars_info): Remove.
10663 (set_reference_optimization_summary): Likewise.
10664 (ipa_init): Create summaries.
10665 (init_function_info): Use function summary.
10666 (ipa_ref_opt_summary_t::duplicate): New function.
10667 (ipa_ref_opt_summary_t::remove): New function.
10668 (get_read_write_all_from_node): Fix GNU coding style.
10669 (propagate): Use function summary.
10670 (write_node_summary_p): Fix GNU coding style.
10671 (stream_out_bitmap): Likewise.
10672 (ipa_reference_read_optimization_summary): Use function summary.
10673 (ipa_reference_c_finalize): Do not release hooks.
10675 2018-06-08 Martin Liska <mliska@suse.cz>
10677 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10678 (analyze_function_body): Extract multiple calls of get_create.
10679 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10680 * ipa-inline.c (recursive_inlining): Use ::get method.
10681 * ipa-inline.h (estimate_edge_growth): Likewise.
10683 2018-06-08 Martin Liska <mliska@suse.cz>
10685 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10687 (hsa_function_summary::hsa_function_summary): Use the new enum
10689 (hsa_gpu_implementation_p): Use hsa_summaries::get.
10690 * hsa-gen.c (hsa_get_host_function): Likewise.
10691 (get_brig_function_name): Likewise.
10692 * ipa-hsa.c (process_hsa_functions): Likewise.
10693 (ipa_hsa_write_summary): Likewise.
10694 * symbol-summary.h (symtab_duplication): Use ::get function/
10695 (get): New function.
10697 2018-06-08 Martin Liska <mliska@suse.cz>
10699 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10701 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10702 (hsa_register_kernel): Likewise.
10703 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10704 * hsa-gen.c (hsa_get_host_function): Likewise.
10705 (get_brig_function_name): Likewise.
10706 (generate_hsa): Likewise.
10707 (pass_gen_hsail::execute): Likewise.
10708 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10709 (devirtualization_time_bonus): Likewise.
10710 (ipcp_propagate_stage): Likewise.
10711 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10712 (edge_set_predicate): Likewise.
10713 (evaluate_conditions_for_known_args): Likewise.
10714 (evaluate_properties_for_edge): Likewise.
10715 (ipa_fn_summary::reset): Likewise.
10716 (ipa_fn_summary_t::duplicate): Likewise.
10717 (dump_ipa_call_summary): Likewise.
10718 (ipa_dump_fn_summary): Likewise.
10719 (analyze_function_body): Likewise.
10720 (compute_fn_summary): Likewise.
10721 (estimate_edge_devirt_benefit): Likewise.
10722 (estimate_edge_size_and_time): Likewise.
10723 (estimate_calls_size_and_time): Likewise.
10724 (estimate_node_size_and_time): Likewise.
10725 (inline_update_callee_summaries): Likewise.
10726 (remap_edge_change_prob): Likewise.
10727 (remap_edge_summaries): Likewise.
10728 (ipa_merge_fn_summary_after_inlining): Likewise.
10729 (ipa_update_overall_fn_summary): Likewise.
10730 (read_ipa_call_summary): Likewise.
10731 (inline_read_section): Likewise.
10732 (write_ipa_call_summary): Likewise.
10733 (ipa_fn_summary_write): Likewise.
10734 (ipa_free_fn_summary): Likewise.
10735 * ipa-hsa.c (process_hsa_functions): Likewise.
10736 (ipa_hsa_write_summary): Likewise.
10737 (ipa_hsa_read_section): Likewise.
10738 * ipa-icf.c (sem_function::merge): Likewise.
10739 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10740 (do_estimate_edge_time): Likewise.
10741 (estimate_size_after_inlining): Likewise.
10742 (estimate_growth): Likewise.
10743 (growth_likely_positive): Likewise.
10744 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10745 (inline_call): Likewise.
10746 * ipa-inline.c (caller_growth_limits): Likewise.
10747 (can_inline_edge_p): Likewise.
10748 (can_inline_edge_by_limits_p): Likewise.
10749 (compute_uninlined_call_time): Likewise.
10750 (compute_inlined_call_time): Likewise.
10751 (want_inline_small_function_p): Likewise.
10752 (edge_badness): Likewise.
10753 (update_caller_keys): Likewise.
10754 (update_callee_keys): Likewise.
10755 (recursive_inlining): Likewise.
10756 (inline_small_functions): Likewise.
10757 (inline_to_all_callers_1): Likewise.
10758 (dump_overall_stats): Likewise.
10759 (early_inline_small_functions): Likewise.
10760 (early_inliner): Likewise.
10761 * ipa-inline.h (estimate_edge_growth): Likewise.
10762 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10763 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10764 * ipa-prop.h (IPA_NODE_REF): Likewise.
10765 (IPA_EDGE_REF): Likewise.
10766 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10767 (propagate_malloc): Likewise.
10768 * ipa-split.c (execute_split_functions): Likewise.
10769 * symbol-summary.h: Rename get to get_create.
10771 (get_create): Likewise.
10772 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10774 2018-06-08 Martin Liska <mliska@suse.cz>
10776 * symbol-summary.h (release): Move definition out of class
10778 (symtab_removal): Likewise.
10779 (symtab_duplication): Likewise.
10781 2018-06-08 Martin Liska <mliska@suse.cz>
10783 * symbol-summary.h (function_summary): Move constructor
10784 implementation out of class declaration.
10785 (release): Likewise.
10786 (symtab_insertion): Likewise.
10787 (symtab_removal): Likewise.
10788 (symtab_duplication): Likewise.
10791 2018-06-08 Martin Liska <mliska@suse.cz>
10793 * Makefile.in: Remove support for MPX (macros, related functions,
10794 fields in cgraph_node, ...).
10795 * builtin-types.def (BT_BND): Likewise.
10796 (BT_FN_BND_CONST_PTR): Likewise.
10797 (BT_FN_CONST_PTR_BND): Likewise.
10798 (BT_FN_VOID_PTR_BND): Likewise.
10799 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10800 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10801 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10802 (expand_builtin_mempcpy_with_bounds): Likewise.
10803 (expand_builtin_memset_with_bounds): Likewise.
10804 (expand_builtin_memset_args): Likewise.
10805 (std_expand_builtin_va_start): Likewise.
10806 (expand_builtin): Likewise.
10807 (expand_builtin_with_bounds): Likewise.
10808 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10809 (DEF_LIB_BUILTIN_CHKP): Likewise.
10810 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10811 (DEF_CHKP_BUILTIN): Likewise.
10812 (BUILT_IN_MEMCPY): Likewise.
10813 (BUILT_IN_MEMMOVE): Likewise.
10814 (BUILT_IN_MEMPCPY): Likewise.
10815 (BUILT_IN_MEMSET): Likewise.
10816 (BUILT_IN_STPCPY): Likewise.
10817 (BUILT_IN_STRCAT): Likewise.
10818 (BUILT_IN_STRCHR): Likewise.
10819 (BUILT_IN_STRCPY): Likewise.
10820 (BUILT_IN_STRLEN): Likewise.
10821 (BUILT_IN_MEMCPY_CHK): Likewise.
10822 (BUILT_IN_MEMMOVE_CHK): Likewise.
10823 (BUILT_IN_MEMPCPY_CHK): Likewise.
10824 (BUILT_IN_MEMSET_CHK): Likewise.
10825 (BUILT_IN_STPCPY_CHK): Likewise.
10826 (BUILT_IN_STRCAT_CHK): Likewise.
10827 (BUILT_IN_STRCPY_CHK): Likewise.
10828 * calls.c (store_bounds): Likewise.
10829 (emit_call_1): Likewise.
10830 (special_function_p): Likewise.
10831 (maybe_warn_nonstring_arg): Likewise.
10832 (initialize_argument_information): Likewise.
10833 (finalize_must_preallocate): Likewise.
10834 (compute_argument_addresses): Likewise.
10835 (expand_call): Likewise.
10836 * cfgexpand.c (expand_call_stmt): Likewise.
10837 (expand_return): Likewise.
10838 (expand_gimple_stmt_1): Likewise.
10839 (pass_expand::execute): Likewise.
10840 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10841 (cgraph_node::remove): Likewise.
10842 (cgraph_node::dump): Likewise.
10843 (cgraph_node::verify_node): Likewise.
10844 * cgraph.h (chkp_function_instrumented_p): Likewise.
10845 (symtab_node::get_alias_target): Likewise.
10846 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10847 (cgraph_local_p): Likewise.
10848 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10849 (cgraph_edge::rebuild_references): Likewise.
10850 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10851 (walk_polymorphic_call_targets): Likewise.
10852 (cgraph_node::expand_thunk): Likewise.
10853 (symbol_table::output_weakrefs): Likewise.
10854 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10855 (ix86_handle_option): Likewise.
10856 * config/i386/constraints.md: Likewise.
10857 * config/i386/i386-builtin-types.def (BND): Likewise.
10861 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10862 (BDESC_FIRST): Likewise.
10864 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10865 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10866 * config/i386/i386.c (enum reg_class): Likewise.
10867 (ix86_target_string): Likewise.
10868 (ix86_option_override_internal): Likewise.
10869 (ix86_conditional_register_usage): Likewise.
10870 (ix86_valid_target_attribute_inner_p): Likewise.
10871 (ix86_set_indirect_branch_type): Likewise.
10872 (ix86_set_current_function): Likewise.
10873 (ix86_function_arg_regno_p): Likewise.
10874 (init_cumulative_args): Likewise.
10875 (ix86_function_arg_advance): Likewise.
10876 (ix86_function_arg): Likewise.
10877 (ix86_pass_by_reference): Likewise.
10878 (ix86_function_value_regno_p): Likewise.
10879 (ix86_function_value_1): Likewise.
10880 (ix86_function_value_bounds): Likewise.
10881 (ix86_return_in_memory): Likewise.
10882 (ix86_setup_incoming_vararg_bounds): Likewise.
10883 (ix86_va_start): Likewise.
10884 (indirect_thunk_need_prefix): Likewise.
10885 (print_reg): Likewise.
10886 (ix86_print_operand): Likewise.
10887 (ix86_expand_call): Likewise.
10888 (ix86_output_function_return): Likewise.
10889 (reg_encoded_number): Likewise.
10890 (BDESC_VERIFYS): Likewise.
10891 (ix86_init_mpx_builtins): Likewise.
10892 (ix86_init_builtins): Likewise.
10893 (ix86_emit_cmove): Likewise.
10894 (ix86_emit_move_max): Likewise.
10895 (ix86_expand_builtin): Likewise.
10896 (ix86_builtin_mpx_function): Likewise.
10897 (ix86_get_arg_address_for_bt): Likewise.
10898 (ix86_load_bounds): Likewise.
10899 (ix86_store_bounds): Likewise.
10900 (ix86_load_returned_bounds): Likewise.
10901 (ix86_store_returned_bounds): Likewise.
10902 (ix86_class_likely_spilled_p): Likewise.
10903 (ix86_hard_regno_mode_ok): Likewise.
10904 (x86_order_regs_for_local_alloc): Likewise.
10905 (ix86_mitigate_rop): Likewise.
10906 (ix86_bnd_prefixed_insn_p): Likewise.
10907 (ix86_mpx_bound_mode): Likewise.
10908 (ix86_make_bounds_constant): Likewise.
10909 (ix86_initialize_bounds): Likewise.
10910 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10911 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10912 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10913 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10914 (TARGET_CHKP_BOUND_MODE): Likewise.
10915 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10916 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10917 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10918 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10919 * config/i386/i386.h (TARGET_MPX): Likewise.
10920 (TARGET_MPX_P): Likewise.
10921 (VALID_BND_REG_MODE): Likewise.
10922 (FIRST_BND_REG): Likewise.
10923 (LAST_BND_REG): Likewise.
10924 (enum reg_class): Likewise.
10925 (BND_REG_P): Likewise.
10926 (BND_REGNO_P): Likewise.
10927 (BNDmode): Likewise.
10928 (ADJUST_INSN_LENGTH): Likewise.
10929 * config/i386/i386.md: Likewise.
10930 * config/i386/i386.opt: Likewise.
10931 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10932 (defined): Likewise.
10933 (LINK_MPX): Likewise.
10934 (MPX_SPEC): Likewise.
10935 (LIBMPX_SPEC): Likewise.
10936 (LIBMPXWRAPPERS_SPEC): Likewise.
10937 (CHKP_SPEC): Likewise.
10938 * config/i386/predicates.md: Likewise.
10939 * dbxout.c (dbxout_type): Likewise.
10940 * doc/extend.texi: Likewise.
10941 * doc/invoke.texi: Likewise.
10942 * doc/md.texi: Likewise.
10943 * doc/tm.texi: Likewise.
10944 * doc/tm.texi.in: Likewise.
10945 * dwarf2out.c (is_base_type): Likewise.
10946 (gen_formal_types_die): Likewise.
10947 (gen_subprogram_die): Likewise.
10948 (gen_type_die_with_usage): Likewise.
10949 (gen_decl_die): Likewise.
10950 (dwarf2out_late_global_decl): Likewise.
10951 * expr.c (expand_assignment): Likewise.
10952 (emit_storent_insn): Likewise.
10953 (store_expr_with_bounds): Likewise.
10954 (store_expr): Likewise.
10955 (expand_expr_real_1): Likewise.
10956 * expr.h (store_expr_with_bounds): Likewise.
10957 * function.c (use_register_for_decl): Likewise.
10958 (struct bounds_parm_data): Likewise.
10959 (assign_parms_augmented_arg_list): Likewise.
10960 (assign_parm_find_entry_rtl): Likewise.
10961 (assign_parm_is_stack_parm): Likewise.
10962 (assign_parm_load_bounds): Likewise.
10963 (assign_bounds): Likewise.
10964 (assign_parms): Likewise.
10965 (expand_function_start): Likewise.
10966 * gcc.c (CHKP_SPEC): Likewise.
10967 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10968 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10969 (wrestrict_dom_walker::check_call): Likewise.
10970 * gimple.c (gimple_build_call_from_tree): Likewise.
10971 * gimple.h (enum gf_mask): Likewise.
10972 (gimple_call_with_bounds_p): Likewise.
10973 (gimple_call_set_with_bounds): Likewise.
10974 * gimplify.c (gimplify_init_constructor): Likewise.
10975 * ipa-cp.c (initialize_node_lattices): Likewise.
10976 (propagate_constants_across_call): Likewise.
10977 (find_more_scalar_values_for_callers_subset): Likewise.
10978 * ipa-hsa.c (process_hsa_functions): Likewise.
10979 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10980 * ipa-icf.c (sem_function::merge): Likewise.
10981 * ipa-inline.c (early_inliner): Likewise.
10982 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10983 (warn_function_cold): Likewise.
10984 (propagate_pure_const): Likewise.
10985 * ipa-ref.h (enum GTY): Likewise.
10986 * ipa-split.c (find_retbnd): Likewise.
10987 (consider_split): Likewise.
10988 (split_function): Likewise.
10989 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10990 * ipa.c (walk_polymorphic_call_targets): Likewise.
10991 (symbol_table::remove_unreachable_nodes): Likewise.
10992 (process_references): Likewise.
10993 (cgraph_build_static_cdtor_1): Likewise.
10994 * lto-cgraph.c (lto_output_node): Likewise.
10995 (output_refs): Likewise.
10996 (compute_ltrans_boundary): Likewise.
10997 (input_overwrite_node): Likewise.
10998 (input_node): Likewise.
10999 (input_cgraph_1): Likewise.
11000 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
11001 * passes.c (pass_manager::execute_early_local_passes): Likewise.
11002 (class pass_chkp_instrumentation_passes): Likewise.
11003 (make_pass_chkp_instrumentation_passes): Likewise.
11004 * passes.def: Likewise.
11005 * rtl.h (struct GTY): Likewise.
11006 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
11007 * stor-layout.c (layout_type): Likewise.
11008 * symtab.c: Likewise.
11009 * target.def: Likewise.
11010 * targhooks.c (default_chkp_bound_type): Likewise.
11011 (default_chkp_bound_mode): Likewise.
11012 (default_builtin_chkp_function): Likewise.
11013 (default_chkp_function_value_bounds): Likewise.
11014 (default_chkp_make_bounds_constant): Likewise.
11015 (default_chkp_initialize_bounds): Likewise.
11016 * targhooks.h (default_chkp_bound_type): Likewise.
11017 (default_chkp_bound_mode): Likewise.
11018 (default_builtin_chkp_function): Likewise.
11019 (default_chkp_function_value_bounds): Likewise.
11020 (default_chkp_make_bounds_constant): Likewise.
11021 (default_chkp_initialize_bounds): Likewise.
11022 * toplev.c (compile_file): Likewise.
11023 (process_options): Likewise.
11024 * tree-core.h (DEF_BUILTIN): Likewise.
11025 (DEF_BUILTIN_CHKP): Likewise.
11026 * tree-inline.c (declare_return_variable): Likewise.
11027 (remap_gimple_stmt): Likewise.
11028 (copy_bb): Likewise.
11029 (initialize_inlined_parameters): Likewise.
11030 (expand_call_inline): Likewise.
11031 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
11032 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
11033 (make_pass_ipa_chkp_produce_thunks): Likewise.
11034 (make_pass_chkp): Likewise.
11035 (make_pass_chkp_opt): Likewise.
11036 (make_pass_chkp_instrumentation_passes): Likewise.
11037 * tree-pretty-print.c (dump_generic_node): Likewise.
11038 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
11039 * tree-ssa-dce.c (propagate_necessity): Likewise.
11040 (eliminate_unnecessary_stmts): Likewise.
11041 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
11042 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
11043 * tree-ssa-sccvn.h: Likewise.
11044 * tree-ssa-strlen.c (get_string_length): Likewise.
11045 (valid_builtin_call): Likewise.
11046 (adjust_last_stmt): Likewise.
11047 (handle_builtin_strchr): Likewise.
11048 (handle_builtin_strcpy): Likewise.
11049 (handle_builtin_stxncpy): Likewise.
11050 (handle_builtin_memcpy): Likewise.
11051 (handle_builtin_strcat): Likewise.
11052 (strlen_check_and_optimize_stmt): Likewise.
11053 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
11054 * tree-streamer-in.c: Likewise.
11055 * tree-streamer.c (record_common_node): Likewise.
11056 * tree.c (tree_code_size): Likewise.
11057 (wide_int_to_tree_1): Likewise.
11058 (type_contains_placeholder_1): Likewise.
11059 (build_common_tree_nodes): Likewise.
11060 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
11061 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
11062 (POINTER_BOUNDS_P): Likewise.
11063 (BOUNDED_TYPE_P): Likewise.
11064 (BOUNDED_P): Likewise.
11065 (CALL_WITH_BOUNDS_P): Likewise.
11066 (pointer_bounds_type_node): Likewise.
11067 * value-prof.c (gimple_ic): Likewise.
11068 * var-tracking.c (vt_add_function_parameters): Likewise.
11069 * varasm.c (make_decl_rtl): Likewise.
11070 (assemble_start_function): Likewise.
11071 (output_constant): Likewise.
11072 (maybe_assemble_visibility): Likewise.
11073 * varpool.c (ctor_for_folding): Likewise.
11074 * chkp-builtins.def: Remove.
11075 * ipa-chkp.c: Remove.
11076 * ipa-chkp.h: Remove.
11077 * rtl-chkp.c: Remove.
11078 * rtl-chkp.h: Remove.
11079 * tree-chkp-opt.c: Remove.
11080 * tree-chkp.c: Remove.
11081 * tree-chkp.h: Remove.
11083 2018-06-07 Carl Love <cel@us.ibm.com>
11085 * config/rs6000/vsx.md (vextract_fp_from_shorth,
11086 vextract_fp_from_shortl): Add BE support.
11088 2018-06-07 Paul Koning <ni1d@arrl.net>
11090 * compare-elim.c (try_merge_compare): Don't merge compare if
11091 address contains a side effect.
11092 (try_eliminate_compare): Likewise.
11094 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
11096 * config.gcc: Support "tremont".
11097 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
11098 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11100 * config/i386/i386.c (m_TREMONT): Define.
11101 (processor_target_table): Add "tremont".
11102 (PTA_TREMONT): Define.
11103 (ix86_lea_outperforms): Add TARGET_TREMONT.
11104 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
11105 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
11106 and M_INTEL_GOLDMONT_PLUS.
11107 (fold_builtin_cpu): Add "tremont".
11108 (ix86_add_stmt_cost): Add TARGET_TREMONT.
11109 (ix86_option_override_internal): Add "tremont".
11110 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
11111 (processor_type): Add PROCESSOR_TREMONT.
11112 * config/i386/x86-tune.def: Add m_TREMONT.
11113 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
11115 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11117 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
11118 symbol defined for msp430i* devices to be lower case.
11120 2018-06-07 Richard Biener <rguenther@suse.de>
11122 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
11123 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
11124 Properly wrap signed arithmetic if overflow wraps.
11126 2018-06-07 Jakub Jelinek <jakub@redhat.com>
11128 PR tree-optimization/69615
11129 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
11130 of a cast from a same precision integral SSA_NAME in a bb dominated
11131 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
11132 cast to utype if rhs2 has already a compatible type.
11134 2018-06-07 Richard Biener <rguenther@suse.de>
11136 PR tree-optimization/85935
11137 * graphite-scop-detection.c (find_params_in_bb): Analyze
11138 condition operands with respect to the correct loop. Assert
11139 the analysis doesn't fail.
11141 2018-06-04 Carl Love <cel@us.ibm.com>
11143 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11144 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
11145 as it is slightly cheaper.
11146 (first_match_or_eos_index_<mode>):
11147 Calculate index using natural element order.
11148 (first_match_index_<mode>):
11149 Calculate index using natural element order.
11150 (first_match_or_eos_index_<mode>):
11151 Calculate index using natural order.
11152 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
11153 for BE and LE modes.
11154 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
11155 P9V_BUILTIN_VCLZLSBB_V16QI.
11156 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
11159 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11161 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
11162 indentation and line wrap for many prototypes. Add missing
11163 @smallexample directives around block of prototypes for vec_xl and
11166 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
11168 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
11169 track if we pass or return IEEE 128-bit floating point.
11170 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
11171 C++ mangling that is compatible with GCC 8.1.
11172 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
11173 (init_cumulative_args): Note if we pass or return IEEE 128-bit
11174 floating point types.
11175 (rs6000_function_arg_advance_1): Likewise.
11176 (rs6000_mangle_type): Optionally generate mangled names that match
11177 what GCC 8.1 generated for IEEE 128-bit floating point types.
11178 (rs6000_globalize_decl_name): If we have an external function that
11179 passes or returns IEEE 128-bit types, generate a weak reference
11180 from the mangled name used in GCC 8.1 to the current mangled
11182 (rs6000_init_builtins): Make __ibm128 use the long double type if
11183 long double is IBM extended double. Make __float128 use the long
11184 double type if long double is IEEE 128-bit.
11187 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
11188 macro for __ibm128 built-in functions.
11189 (PACK_IF): Add __ibm128 pack/unpack functions.
11190 (UNPACK_IF): Likewise.
11191 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
11192 enable long double built-in functions if long double is IEEE
11193 128-bit floating point.
11194 (rs6000_invalid_builtin): Update long double built-in function
11196 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
11197 functions, adjust the built-in function to use the long double
11198 built-in function if __ibm128 and long double are the same type.
11199 * doc/extend.texi (PowerPC builtins): Update documention for
11200 __builtin_{,un}pack_longdouble. Add documentation for
11201 __builtin_{,un}pack_ibm128.
11203 2018-06-06 Jim Wilson <jimw@sifive.com>
11205 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
11206 (struct machine_function): New field interrupt_mode.
11207 (riscv_handle_type_attribute): New function. Add forward declaration.
11208 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
11209 (riscv_expand_epilogue): Check interrupt_mode field.
11210 (riscv_set_current_function): Check interrupt attribute args and
11211 set interrupt_mode field.
11212 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
11213 (riscv_sret, riscv_uret): New.
11214 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
11215 new arguments to interrupt attribute.
11217 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
11220 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
11221 Don't handle -mcpu=power8 if -mpower9-vector is also used.
11223 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11225 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
11226 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11227 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11228 several redundant entries.
11230 2018-06-06 David Malcolm <dmalcolm@redhat.com>
11232 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11233 type from "rtx" to "rtx_insn *".
11234 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
11235 for local "call_insn", removing cast.
11236 (ix86_expand_call): Likewise, introducing a "call_insn" local.
11238 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
11240 PR tree-optimization/86066
11241 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
11242 for BIT_INSERT_EXPR stores.
11244 2018-06-06 Richard Biener <rguenther@suse.de>
11246 PR tree-optimization/86062
11247 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11248 component refs ontop
11249 of to be offsetted base.
11251 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11253 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
11254 to be static and remove check on interrupt attribute name.
11256 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
11258 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11259 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
11261 2018-06-05 Steve Ellcey <sellcey@cavium.com>
11264 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11266 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
11267 Remove second argument, change how error is called.
11268 (aarch64_layout_arg): Remove second argument from
11269 aarch64_err_no_fpadvsimd call.
11270 (aarch64_init_cumulative_args): Ditto.
11271 (aarch64_gimplify_va_arg_expr): Ditto.
11272 * config/aarch64/aarch64.md (mov<mode>): Ditto.
11274 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
11276 * config/i386/i386.md (simple_return_indirect_internal): New expander.
11277 (*simple_return_indirect_internal<mode>): Rename from
11278 simple_return_indirect_internal. Use W mode iterator.
11279 (rstorssp): New expander.
11280 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
11281 (clrssbsy): New expander.
11282 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
11284 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
11286 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
11288 (cmse_check_pointed_object): Likewise.
11290 2018-06-05 Martin Liska <mliska@suse.cz>
11292 PR gcov-profile/47618
11293 * doc/invoke.texi: Document how -fprofile-dir format
11296 2018-06-05 Richard Biener <rguenther@suse.de>
11298 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
11299 removal pretend DOM info isn't available so we do not update
11300 it and only remove edges, not dominated blocks. Actually free
11301 DOM info in case we removed something. Remove unreachable blocks.
11302 (mfb_keep_latches): Work with either DOM info or marked backedges.
11303 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
11304 first. Mark backedges if DOM info isn't available.
11305 (Re-)compute DOM info after cleanup_control_flow_pre.
11307 2018-06-05 Richard Biener <rguenther@suse.de>
11309 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
11310 (locus_discrim_hasher::hash): Adjust.
11311 (locus_discrim_hasher::equal): Likewise.
11312 (next_discriminator_for_locus): Work on line directly.
11313 (same_line_p): Pass in expanded locus1 as well.
11314 (assign_discriminators): Avoid redundant location expansions.
11316 2018-06-05 Richard Biener <rguenther@suse.de>
11318 PR tree-optimization/86046
11319 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
11320 if required after clearing TREE_ADDRESSABLE.
11322 2018-06-05 Richard Biener <rguenther@suse.de>
11324 PR tree-optimization/86047
11325 * tree-ssa-loop.c (for_each_index): Glob handling of all
11326 decls and constants and really handle all of them.
11328 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11331 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
11332 qualifier_void_pointer and qualifier_const_void_pointer.
11333 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
11334 (arm_init_builtins): Handle the above.
11335 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
11336 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
11337 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
11340 2018-06-05 Martin Liska <mliska@suse.cz>
11342 * auto-profile.c (read_autofdo_file): Do not use
11343 gcov_ctr_summary struct.
11344 (afdo_callsite_hot_enough_for_early_inline): Likewise.
11345 * coverage.c (struct counts_entry): Likewise.
11346 (read_counts_file): Read just single summary entry.
11347 (get_coverage_counts): Use gcov_summary struct.
11348 * coverage.h (get_coverage_counts): Likewise.
11349 * gcov-dump.c (dump_working_sets): Likewise.
11350 (tag_summary): Dump just single summary.
11351 * gcov-io.c (gcov_write_summary): Write just histogram
11353 (gcov_read_summary): Read just single summary.
11354 (compute_working_sets): Use gcov_summary struct.
11355 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
11356 of GCOV_COUNTERS_SUMMABLE.
11357 (GCOV_COUNTERS_SUMMABLE): Remove.
11358 (GCOV_FIRST_VALUE_COUNTER): Replace with
11359 GCOV_COUNTER_V_INTERVAL.
11360 (struct gcov_ctr_summary): Remove.
11361 (struct gcov_summary): Directly use fields of former
11363 (compute_working_sets): Use gcov_summary struct.
11364 * gcov.c (read_count_file): Do not use ctrs fields.
11365 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
11367 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
11369 * profile.c: Likewise.
11370 * profile.h: Likewise.
11372 2018-06-05 Martin Liska <mliska@suse.cz>
11374 PR gcov-profile/84846
11375 * gcov.c (output_lines): Print working directory only
11376 in intermediate format.
11378 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
11380 * config/s390/s390-builtin-types.def: Add void function type.
11381 * config/s390/s390-builtins.def: Use the function type for the
11384 2018-06-04 Jim Wilson <jimw@sifive.com>
11386 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
11388 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
11389 and maybe_eh_return. Change regno to unsigned int. Use new args to
11390 handle EH_RETURN_DATA_REGNO registers properly.
11391 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
11392 (riscv_expand_epilogue): Update comment. Change argument name and
11393 type. Update code to use new name and type. Pass new args to
11394 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
11396 * config/riscv/riscv.md (NORMAL_RETURN): New.
11397 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
11398 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
11399 (eh_return): Call gen_eh_return_internal and emit barrier.
11400 (eh_return_internal): Call riscv_expand_epilogue.
11402 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
11404 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
11405 bit_insertion field and declare can_be_merged_into method.
11406 (merged_store_group::can_be_merged_into): New method.
11407 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
11408 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
11409 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
11411 2018-06-04 Richard Biener <rguenther@suse.de>
11413 PR tree-optimization/85955
11414 * builtins.c (fold_builtin_sincos): Convert pointers to
11415 destination to appropriate type before dereferencing.
11417 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
11419 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
11421 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
11423 * expr.c (expand_expr_real_1): Force the operand into memory if
11424 its TYPE_MODE is BLKmode and if there is no integer mode for
11425 the number of bits being extracted.
11427 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11431 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11432 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
11434 2018-06-04 Richard Biener <rguenther@suse.de>
11436 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
11437 (cleanup_tree_cfg_noloop): ... single caller. Do
11438 start_recording_case_labels later.
11440 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
11442 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11443 to _IMMINTRIN_H_INCLUDED.
11444 * config/i386/pconfigintrin.h: Ditto.
11445 * config/i386/waitpkgintrin.h: Ditto.
11446 * config/i386/immintrin.h: Add includes for sgxintrin.h,
11447 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11448 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11449 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11450 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11451 waitpkgintrin.h and cldemoteintrin.h.
11453 2018-06-04 Richard Biener <rguenther@suse.de>
11455 PR tree-optimization/86038
11456 * tracer.c (find_best_successor): Check probability for
11457 being initialized, bail out if not.
11459 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
11462 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11463 of bits to ignore when comparing architectures.
11465 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11467 PR tree-optimization/69615
11468 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11469 maximum or minimum of the type, try to merge it also as if
11470 range1 is + [-, x - 1] or + [x + 1, -].
11473 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11475 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
11477 PR tree-optimization/86034
11478 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11479 the unsigned bitfield type in a bit insertion sequence if it does not
11480 have a larger precision than the bitfield size.
11481 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11483 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11485 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11487 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11489 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11490 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11491 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11492 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11494 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11496 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11497 Disable -fdelete-null-pointer-checks for ELF toolchain.
11499 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11500 Kito Cheng <kito.cheng@gmail.com>
11502 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11503 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11504 (nds32*-*-*): Add glibc and uclibc conditions.
11505 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11506 (TARGET_EXCEPT_UNWIND_INFO): Define.
11507 * config/nds32/elf.h: New file.
11508 * config/nds32/linux.h: New file.
11509 * config/nds32/nds32-elf.opt: New file.
11510 * config/nds32/nds32-linux.opt: New file.
11511 * config/nds32/nds32-fp-as-gp.c
11512 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11513 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11515 (nds32_asm_file_end): Ditto.
11516 (nds32_print_operand): Ditto.
11517 (nds32_insert_attributes): Ditto.
11518 (nds32_init_libfuncs): New function.
11519 (TARGET_HAVE_TLS): Define.
11520 (TARGET_INIT_LIBFUNCS): Define.
11521 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11523 (TARGET_ELF): Apply different mcmodel setting.
11524 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11525 been migrated into elf.h and linux.h files.
11526 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11527 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11528 (mcmodel): The content has been migrated into nds32-elf.opt and
11529 nds32-linux.opt files.
11530 * config/nds32/t-elf: New file.
11531 * config/nds32/t-linux: New file.
11533 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11534 Shiva Chen <shiva0217@gmail.com>
11536 * config/nds32/constants.md (unspec_volatile_element): Add
11537 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11538 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11540 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11541 (make_pass_nds32_fp_as_gp): Declare.
11542 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11544 (nds32_asm_function_end_prologue): Remove unused asm output.
11545 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11546 (nds32_asm_file_start): Output necessary fp_as_gp information.
11547 (nds32_option_override): Adjust register usage.
11548 (nds32_expand_prologue): Consider fp_as_gp situation.
11549 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11550 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11553 (simple_return): Ditto.
11554 (omit_fp_begin): Output special directive for fp_as_gp.
11555 (omit_fp_end): Output special directive for fp_as_gp.
11556 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11557 mforbid-fp-as-gp): New options.
11559 2018-06-01 Mark Wielaard <mark@klomp.org>
11561 * dwarf2out.c (dwarf2out_finish): Remove generation of
11562 DW_AT_loclists_base.
11564 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11566 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11567 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11568 (struct merged_store_group): Add bit_insertion field.
11569 (dump_char_array): Use standard hexadecimal format.
11570 (merged_store_group::merged_store_group): Set bit_insertion to false.
11571 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11572 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11573 also print the mask in the dump file.
11574 (pass_store_merging::gate): Minor tweak.
11575 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11576 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11577 stores with INTEGER_CST stores.
11578 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11579 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11580 and use it throughout. Generate bit insertion sequences if need be.
11581 (pass_store_merging::process_store): Remove redundant condition.
11582 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11584 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11586 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11587 the 128-bit floating point types. Fix function comment.
11589 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11591 * config/aarch64/aarch64-simd.md
11592 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11594 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11597 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11599 PR tree-optimization/85989
11600 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11602 (backprop::intersect_uses): Check it when deciding whether this
11603 is a backedge reference.
11604 (backprop::process_block): Add each phi to m_visited_phis
11605 after visiting it, then clear it at the end.
11607 2018-06-01 Richard Biener <rguenther@suse.de>
11609 * tree-vectorizer.h (vect_dr_stmt): New function.
11610 (vect_get_load_cost): Adjust.
11611 (vect_get_store_cost): Likewise.
11612 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11613 Use vect_dr_stmt instead of DR_SMTT.
11614 (vect_record_base_alignments): Likewise.
11615 (vect_calculate_target_alignment): Likewise.
11616 (vect_compute_data_ref_alignment): Likewise and make static.
11617 (vect_update_misalignment_for_peel): Likewise.
11618 (vect_verify_datarefs_alignment): Likewise.
11619 (vector_alignment_reachable_p): Likewise.
11620 (vect_get_data_access_cost): Likewise. Pass down
11621 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11622 (vect_get_peeling_costs_all_drs): Likewise.
11623 (vect_peeling_hash_get_lowest_cost): Likewise.
11624 (vect_enhance_data_refs_alignment): Likewise.
11625 (vect_find_same_alignment_drs): Likewise.
11626 (vect_analyze_data_refs_alignment): Likewise.
11627 (vect_analyze_group_access_1): Likewise.
11628 (vect_analyze_group_access): Likewise.
11629 (vect_analyze_data_ref_access): Likewise.
11630 (vect_analyze_data_ref_accesses): Likewise.
11631 (vect_vfa_segment_size): Likewise.
11632 (vect_small_gap_p): Likewise.
11633 (vectorizable_with_step_bound_p): Likewise.
11634 (vect_prune_runtime_alias_test_list): Likewise.
11635 (vect_analyze_data_refs): Likewise.
11636 (vect_supportable_dr_alignment): Likewise.
11637 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11638 (vect_gen_prolog_loop_niters): Likewise.
11639 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11640 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11642 (vect_recog_mask_conversion_pattern): Likewise.
11643 (vect_try_gather_scatter_pattern): Likewise.
11644 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11645 to vect_get_store_cost.
11646 (vect_get_store_cost): Get stmt_info instead of DR.
11647 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11648 (vect_get_load_cost): Get stmt_info instead of DR.
11650 2018-06-01 Richard Biener <rguenther@suse.de>
11652 PR middle-end/86017
11653 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11654 vars wrapped in MEM_REFs.
11656 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11658 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11659 Fix subreg tests so that we only return a choice between
11660 GENERAL_REGS and FP_REGS if the original classes included both.
11662 2018-06-01 Richard Biener <rguenther@suse.de>
11665 * tree-ssa-structalias.c (get_function_part_constraint):
11666 Handle NULL fi->decl.
11667 (find_func_aliases_for_call): Properly handle indirect
11668 fi from direct call.
11669 (find_func_clobbers): Likewise.
11670 (ipa_pta_execute): Likewise.
11671 (create_variable_info_for): For functions that are ifunc_resolver
11672 resolve to a varinfo that contains the result of the resolver call.
11673 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11676 2018-05-31 Michael Collison <michael.collison@arm.com>
11678 * config/aarch64/aarch64.md:
11679 (*fix_to_zero_extenddfdi2): New pattern.
11680 * gcc.target/aarch64/fix_extend1.c: New testcase.
11682 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
11684 PR middle-end/78809
11685 PR middle-end/83026
11686 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11687 and BUILT_IN_STRNCMP_EQ.
11688 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11689 BUILT_IN_STRNCMP_EQ.
11690 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11691 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11692 (gimple_fold_builtin): Likewise.
11693 * tree-ssa-strlen.c (compute_string_length): New function.
11694 (determine_min_obsize): New function.
11695 (handle_builtin_string_cmp): New function to handle calls to
11696 string compare functions.
11697 (strlen_optimize_stmt): Add handling to builtin string compare
11699 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11700 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11701 * tree.c (build_common_builtin_nodes): Add new defines of
11702 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11704 2018-05-31 Jakub Jelinek <jakub@redhat.com>
11707 * bb-reorder.c (pass_partition_blocks::gate): Return false for
11708 functions with naked attribute.
11710 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
11712 * config/i386/sse.md (avx_vec_concat<mode>):
11713 Substitute concat_tg_mode mode attribute with xtg_mode.
11714 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11715 (concat_tg_mode): Remove mode attribute.
11717 2018-05-31 Martin Sebor <msebor@redhat.com>
11720 * calls.c (alloc_max_size): Correct a logic error/typo.
11721 Treat excessive arguments as infinite. Warn for invalid arguments.
11722 * doc/invoke.texi (-Walloc-size-larger-than): Update.
11724 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
11727 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11728 and movx for Haswell.
11730 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
11731 Cesar Philippidis <cesar@codesourcery.com>
11733 PR middle-end/85879
11734 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11735 when emitting error on private/firstprivate reductions.
11736 * omp-low.c (lower_omp_target): Avoid reference-type processing
11737 on pointers for firstprivate clause.
11739 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11741 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11744 * config/aarch64/aarch64-simd.md
11745 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11746 (aarch64_ld1_x3_<mode>): Likewise
11747 (aarch64_st1x2<VALLDIF:mode>): Likewise
11748 (aarch64_st1_x2_<mode>): Likewise
11749 (aarch64_st1x3<VALLDIF:mode>): Likewise
11750 (aarch64_st1_x3_<mode>): Likewise
11751 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11752 (vld1_s8_x3): Likewise.
11753 (vld1_u16_x3): Likewise.
11754 (vld1_s16_x3): Likewise.
11755 (vld1_u32_x3): Likewise.
11756 (vld1_s32_x3): Likewise.
11757 (vld1_u64_x3): Likewise.
11758 (vld1_s64_x3): Likewise.
11759 (vld1_f16_x3): Likewise.
11760 (vld1_f32_x3): Likewise.
11761 (vld1_f64_x3): Likewise.
11762 (vld1_p8_x3): Likewise.
11763 (vld1_p16_x3): Likewise.
11764 (vld1_p64_x3): Likewise.
11765 (vld1q_u8_x3): Likewise.
11766 (vld1q_s8_x3): Likewise.
11767 (vld1q_u16_x3): Likewise.
11768 (vld1q_s16_x3): Likewise.
11769 (vld1q_u32_x3): Likewise.
11770 (vld1q_s32_x3): Likewise.
11771 (vld1q_u64_x3): Likewise.
11772 (vld1q_s64_x3): Likewise.
11773 (vld1q_f16_x3): Likewise.
11774 (vld1q_f32_x3): Likewise.
11775 (vld1q_f64_x3): Likewise.
11776 (vld1q_p8_x3): Likewise.
11777 (vld1q_p16_x3): Likewise.
11778 (vld1q_p64_x3): Likewise.
11779 (vst1_s64_x2): Likewise.
11780 (vst1_u64_x2): Likewise.
11781 (vst1_f64_x2): Likewise.
11782 (vst1_s8_x2): Likewise.
11783 (vst1_p8_x2): Likewise.
11784 (vst1_s16_x2): Likewise.
11785 (vst1_p16_x2): Likewise.
11786 (vst1_s32_x2): Likewise.
11787 (vst1_u8_x2): Likewise.
11788 (vst1_u16_x2): Likewise.
11789 (vst1_u32_x2): Likewise.
11790 (vst1_f16_x2): Likewise.
11791 (vst1_f32_x2): Likewise.
11792 (vst1_p64_x2): Likewise.
11793 (vst1q_s8_x2): Likewise.
11794 (vst1q_p8_x2): Likewise.
11795 (vst1q_s16_x2): Likewise.
11796 (vst1q_p16_x2): Likewise.
11797 (vst1q_s32_x2): Likewise.
11798 (vst1q_s64_x2): Likewise.
11799 (vst1q_u8_x2): Likewise.
11800 (vst1q_u16_x2): Likewise.
11801 (vst1q_u32_x2): Likewise.
11802 (vst1q_u64_x2): Likewise.
11803 (vst1q_f16_x2): Likewise.
11804 (vst1q_f32_x2): Likewise.
11805 (vst1q_f64_x2): Likewise.
11806 (vst1q_p64_x2): Likewise.
11807 (vst1_s64_x3): Likewise.
11808 (vst1_u64_x3): Likewise.
11809 (vst1_f64_x3): Likewise.
11810 (vst1_s8_x3): Likewise.
11811 (vst1_p8_x3): Likewise.
11812 (vst1_s16_x3): Likewise.
11813 (vst1_p16_x3): Likewise.
11814 (vst1_s32_x3): Likewise.
11815 (vst1_u8_x3): Likewise.
11816 (vst1_u16_x3): Likewise.
11817 (vst1_u32_x3): Likewise.
11818 (vst1_f16_x3): Likewise.
11819 (vst1_f32_x3): Likewise.
11820 (vst1_p64_x3): Likewise.
11821 (vst1q_s8_x3): Likewise.
11822 (vst1q_p8_x3): Likewise.
11823 (vst1q_s16_x3): Likewise.
11824 (vst1q_p16_x3): Likewise.
11825 (vst1q_s32_x3): Likewise.
11826 (vst1q_s64_x3): Likewise.
11827 (vst1q_u8_x3): Likewise.
11828 (vst1q_u16_x3): Likewise.
11829 (vst1q_u32_x3): Likewise.
11830 (vst1q_u64_x3): Likewise.
11831 (vst1q_f16_x3): Likewise.
11832 (vst1q_f32_x3): Likewise.
11833 (vst1q_f64_x3): Likewise.
11834 (vst1q_p64_x3): Likewise.
11836 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11838 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11839 user_label_prefix to name.
11841 * tree-core.h: Update comment about the format of NAME string
11842 passed to handler in attribute_spec.
11844 * config/msp430/msp430.md: Remove erroneous subreg expression from
11845 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11846 zero_extend{q,h}isi2.
11848 2018-05-30 Borislav Petkov <bp@suse.de>
11850 * doc/extend.texi: Document some architecture specific
11851 constraints and sort entries.
11853 2018-05-30 Martin Sebor <msebor@redhat.com>
11855 PR middle-end/85369
11856 * builtins.c (expand_builtin_stpcpy_1): New function.
11857 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11858 only if the former succeeds.
11860 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11862 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11865 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11867 * doc/invoke.texi (-flinker-output): Document
11869 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11871 * passes.c (ipa_write_summaries): Only modify statements if body
11873 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11874 incrementally linking.
11875 (ipa_passes): Likewise.
11876 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11877 pass down resolution info.
11878 * common.opt (flag_incremental_link): Update info.
11879 * gcc.c (plugin specs): Turn flinker-output=* to
11880 -plugin-opt=-linker-output-known
11881 * toplev.c (compile_file): Also cut compilation when doing incremental
11883 * flag-types. (enum lto_partition_model): Add
11884 LTO_LINKER_OUTPUT_NOLTOREL.
11885 (invoke.texi): Add -flinker-output docs.
11886 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11887 link same way as WPA; do not stream in dead initializers.
11889 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11890 darf2out_register_external_decl): Support incremental link.
11892 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11894 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11896 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11898 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11899 it down to simple_object_copy_lto_debug_sections.
11900 (run_gcc): Determine incremental LTO link time and configure
11901 lto1 into non-wpa mode, disable renaming of debug sections.
11903 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11905 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11906 descriptions of various incorrectly documented functions.
11908 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11911 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11912 address check not strict.
11914 2018-05-30 Richard Biener <rguenther@suse.de>
11916 PR tree-optimization/85964
11917 * tracer.c (better_p): Drop initialized count check, we only
11918 call the function with initialized counts now.
11919 (find_best_successor): Do find a best edge if one
11920 has uninitialized count.
11921 (find_best_predecessor): Likewise. Do BB frequency check only
11922 if count is initialized.
11924 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11926 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11927 (aarch64_ldrstr_offset_compare): New.
11928 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11929 load/store orderings.
11930 (aarch64_gen_adjusted_ldpstp): Likewise.
11932 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11934 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11935 Check for subset of GENERAL_REGS and FP_REGS.
11936 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11939 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11941 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11942 and wi::to_poly_offset. Add the current offset and then check
11943 whether the sum fits, rather than using an unchecked addition of
11944 a checked term. Check for a shwi rather than a uhwi.
11945 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11946 (store_constructor): Use poly_int_tree_p.
11947 (expand_expr_real_1): Likewise.
11948 * function.c (assign_temp): Likewise.
11949 * fold-const.c (const_binop): Use poly_int_tree_p and
11950 wi::to_poly_offset.
11951 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11953 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11954 to_poly_offset for MEM offsets.
11955 * ipa-icf.c (sem_variable::equals): Likewise.
11956 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11957 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11958 wi::to_poly_offset for BIT_FIELD_REF offsets.
11959 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11960 wi::to_poly_offset.
11961 * var-tracking.c (emit_note_insn_var_location): Use
11962 tree_to_poly_uint64.
11964 2018-05-29 Jim Wilson <jimw@sifive.com>
11966 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11968 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11971 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11972 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11974 (sse4_1_round<mode>2): Use nonimmediate_operand
11975 for operand 1 predicate.
11977 2018-05-29 Martin Sebor <msebor@redhat.com>
11978 Richard Biener <rguenther@suse.de>
11981 * calls.c (get_size_range): Call determine_value_range instead
11982 of get_value_range..
11983 * tree-vrp.h (determine_value_range): Declared new function.
11984 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11986 2018-05-29 Richard Biener <rguenther@suse.de>
11988 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11989 sure to use non-pattern stmts for get_earlier_stmt arguments.
11990 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11991 called on pattern stmts.
11992 (get_later_stmt): Likewise.
11994 2018-05-29 Martin Liska <mliska@suse.cz>
11996 PR gcov-profile/85759
11997 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
12000 2018-05-29 Jakub Jelinek <jakub@redhat.com>
12002 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
12004 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
12008 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12009 VEC_PACK_FLOAT_EXPR): New tree codes.
12010 * tree-pretty-print.c (op_code_prio): Handle
12011 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12012 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12013 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12014 * tree-inline.c (estimate_operator_cost): Likewise.
12015 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
12016 * fold-const.c (const_binop): Likewise.
12017 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
12018 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12019 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12020 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
12021 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12022 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12023 * expr.c (expand_expr_real_2): Likewise.
12024 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
12025 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
12026 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
12028 * optabs.c (expand_widen_pattern_expr): For
12029 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
12030 sign from result type rather than operand's type.
12031 (expand_binop_directly): For vec_packu_float_optab and
12032 vec_packs_float_optab allow result type to be different from operand's
12034 * optabs-tree.c (optab_for_tree_code): Handle
12035 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12036 VEC_PACK_FLOAT_EXPR. Formatting fixes.
12037 * tree-vect-generic.c (expand_vector_operations_1): Handle
12038 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12039 VEC_PACK_FLOAT_EXPR.
12040 * tree-vect-stmts.c (supportable_widening_operation): Handle
12042 (supportable_narrowing_operation): Handle FLOAT_EXPR.
12043 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
12044 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
12045 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
12046 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
12048 (vec_pack<floatprefix>_float_<mode>): New expander.
12049 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
12051 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
12052 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
12053 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
12054 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
12055 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
12057 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
12058 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
12059 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12060 VEC_PACK_FLOAT_EXPR): Document.
12062 2018-05-29 Richard Biener <rguenther@suse.de>
12064 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
12066 (stmt_vec_info_vec): Make pointer.
12067 (init_stmt_vec_info_vec): Remove.
12068 (free_stmt_vec_info_vec): Likewise.
12069 (set_stmt_vec_info_vec): New function.
12070 (free_stmt_vec_infos): Likewise.
12071 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
12072 (set_vinfo_for_stmt): Likewise.
12073 (get_earlier_stmt): Likewise.
12074 (get_later_stmt): Likewise.
12075 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
12076 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
12077 (vec_info::~vec_info): Free stmt_vec_infos.
12078 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
12079 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
12080 (pass_slp_vectorize::execute): Likewise.
12081 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
12082 (free_stmt_vec_info_vec): Likewise.
12083 (set_stmt_vec_info_vec): New function.
12084 (free_stmt_vec_infos): Likewise.
12085 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
12086 the global stmt_vec_info_vec.
12087 * tree-parloops.c (gather_scalar_reductions): Use
12088 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
12091 2018-05-29 Richard Biener <rguenther@suse.de>
12093 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
12095 2018-05-29 Martin Liska <mliska@suse.cz>
12096 David Malcolm <dmalcolm@redhat.com>
12098 * vec.c (test_reverse): New.
12099 (vec_c_tests): Add new test.
12100 * vec.h (vl_ptr>::reverse): New function.
12102 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
12104 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
12106 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
12109 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
12111 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
12113 2018-05-28 Richard Biener <rguenther@suse.de>
12115 PR tree-optimization/85933
12116 * tree-vect-data-refs.c (vect_record_base_alignments): Only
12117 look at stmts marked as vectorizable.
12119 2018-05-28 Richard Biener <rguenther@suse.de>
12121 PR tree-optimization/85934
12122 * tree-vect-generic.c (expand_vector_operations_1): Hoist
12123 vector boolean check before scalar optimization.
12125 2018-05-28 Jakub Jelinek <jakub@redhat.com>
12127 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
12130 2018-05-28 Mark Wielaard <mark@klomp.org>
12132 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
12133 if it is an expression containing a minus sign.
12135 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
12137 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
12139 2018-05-27 Paul Koning <ni1d@arrl.net>
12141 * config/pdp11/pdp11.md (truncsihi2): Remove.
12143 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
12144 Chung-Ju Wu <jasonwucj@gmail.com>
12146 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
12148 (unaligned_store_dw): Ditto.
12149 * config/nds32/nds32-memory-manipulation.c
12150 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
12151 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
12152 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
12153 (emit_setmem_word_loop): Rename to ...
12154 (emit_setmem_doubleword_loop): ... this.
12155 (nds32_gen_dup_4_byte_to_word_value): New function.
12156 (nds32_gen_dup_8_byte_to_double_word_value): New function.
12157 (nds32_expand_setmem_loop): Refine implementation.
12158 (nds32_expand_setmem_loop_v3m): Ditto.
12159 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
12162 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12164 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
12166 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12168 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
12169 (nds32_init_machine_status): Initialize machine->attr_naked_p and
12170 machine->attr_no_prologue_p.
12171 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
12172 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12173 (nds32_expand_epilogue): Consider attr_naked_p.
12174 (nds32_expand_epilogue_v3pop): Likewise.
12175 (nds32_can_use_return_insn): Likewise.
12176 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12177 attr_no_prologue_p fields.
12178 * config/nds32/nds32.opt (mret-in-naked-func): New option.
12180 2018-05-27 Jakub Jelinek <jakub@redhat.com>
12183 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
12185 * config/i386/sse.md
12186 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
12188 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
12190 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
12192 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
12194 (*<floatsuffix>floatv2div2sf2): Rename to ...
12195 (*float<floatunssuffix>v2div2sf2): ... this.
12196 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
12197 (float<floatunssuffix>v2div2sf2_mask): ... this.
12198 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
12199 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
12200 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
12202 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
12204 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12206 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12208 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12210 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12212 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
12213 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
12214 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
12215 gen_ufix_truncv8dfv8si2.
12216 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
12217 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
12218 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
12219 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
12220 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
12221 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
12222 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
12223 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
12225 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12229 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
12231 2018-05-25 Jim Wilson <jimw@sifive.com>
12233 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
12234 * config/riscv/riscv.c (struct machine_function): Add
12235 interrupt_handler_p and attribute_checked_p fields.
12236 (riscv_attribute_table): Add interrupt.
12237 (riscv_interrupt_type_p): New.
12238 (riscv_save_reg_p): Save extra regs for interrupt handler.
12239 (riscv_use_save_libcall): Return false for interrupt handler.
12240 (riscv_first_stack_step): Add forward declaration.
12241 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
12242 for interrupt handler with large frame. Use it for saved reg list.
12243 (riscv_expand_prologue): Move flag_stack_usage_info support to
12244 eliminate duplication.
12245 (riscv_expand_epilogue): Generate mret for interrupt handler.
12246 (riscv_epilogue_uses): New.
12247 (riscv_can_use_return_insn): Return false for interrupt handler.
12248 (riscv_function_ok_for_sibcall): Likewise.
12249 (riscv_set_current_function): Add interrupt handler support.
12250 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
12251 * config/riscv/riscv.md (UNSPECV_MRET): New.
12253 (riscv_frflags, riscv_fsflags): Use tab after opcode.
12255 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
12257 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
12259 PR tree-optimization/85712
12260 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
12261 this candidate has already been replaced in-situ by a copy.
12263 2018-05-25 Jason Merrill <jason@redhat.com>
12265 PR c++/80485 - inline function non-zero address.
12266 * symtab.c (nonzero_address): Check DECL_COMDAT.
12268 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
12271 * config/alpha/alpha.md (ashlsi3): New insn pattern.
12272 (*ashlsi_se): Rename from *ashldi_se. Define as sign
12273 extension of SImode operation. Use const123_operand predicate.
12274 (*saddsi_1): Remove.
12275 (*saddl_se_1): Ditto.
12276 (*ssubsi_1): Ditto.
12277 (*ssubl_se_1): Ditto.
12278 * config/alpha/predicates.md (const123_operand): New predicate.
12279 * config/alpha/constraints.md (P): Use IN_RANGE.
12281 2018-05-25 Richard Biener <rguenther@suse.de>
12283 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
12285 (ref_maybe_used_by_stmt_p): Likewise.
12286 (stmt_may_clobber_ref_p): Likewise.
12287 (stmt_may_clobber_ref_p_1): Likewise.
12288 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
12290 (ref_maybe_used_by_stmt_p): Likewise.
12291 (stmt_may_clobber_ref_p): Likewise.
12292 (stmt_may_clobber_ref_p_1): Likewise.
12293 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
12294 the alias oracle to disambiguate DRs with stmts DR analysis
12296 (vect_analyze_data_refs): Do not give up on not analyzable
12297 DRs for BB vectorization. Remove code truncating the dataref
12300 2018-05-25 Jakub Jelinek <jakub@redhat.com>
12303 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12304 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
12305 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
12307 2018-05-25 Richard Biener <rguenther@suse.de>
12309 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
12310 function, combining stmt data ref gathering and fatal analysis
12312 (vect_analyze_data_refs): Remove now redudnant code and simplify.
12313 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
12314 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
12315 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
12316 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
12318 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12320 PR tree-optimization/85720
12321 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
12322 SCC if all partitions are builtins.
12323 (version_loop_by_alias_check): New parameter. Generate cancelable
12324 runtime alias check if all partitions are builtins.
12325 (distribute_loop): Update call to above function.
12327 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12329 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
12330 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
12331 (parm_default_def_partition_arg): Ditto.
12332 (set_parm_default_def_partition): Ditto.
12333 (get_parm_default_def_partitions): Ditto and make it static.
12334 (get_undefined_value_partitions): Ditto and make it static.
12335 (remove_ssa_form): Refactor call to init_var_map here.
12336 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
12337 computation for loop region.
12338 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
12339 (register_default_def): Delete.
12340 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
12341 (parm_default_def_partition_arg): Ditto.
12342 (set_parm_default_def_partition): Ditto.
12343 (get_parm_default_def_partitions): Ditto and make it static.
12344 (get_undefined_value_partitions): Ditto and make it static.
12345 (coalesce_with_default, coalesce_with_default): Update comment.
12346 (create_coalesce_list_for_region): New func factored out from
12347 create_outofssa_var_map.
12348 (populate_coalesce_list_for_outofssa): New func factored out from
12349 create_outofssa_var_map and coalesce_ssa_name.
12350 (create_outofssa_var_map): Delete.
12351 (coalesce_ssa_name): Refactor to support live range computation.
12352 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
12353 (get_parm_default_def_partitions): Delete.
12354 (get_undefined_value_partitions): Ditto.
12355 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
12356 computation for loop region.
12357 (new_tree_live_info, loe_visit_block): Ditto.
12358 (live_worklist, set_var_live_on_entry): Ditto.
12359 (calculate_live_on_exit, verify_live_on_entry): Ditto.
12360 * tree-ssa-live.h (struct _var_map): New fields.
12361 (init_var_map): Change decl.
12362 (region_contains_p): New.
12364 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12366 * tree-ssa-live.h (live_merge_and_clear): Delete.
12368 2018-05-25 Richard Biener <rguenther@suse.de>
12371 * tree-dump.c (dequeue_and_dump): Remove access to removed
12372 operand 2 of a SWITCH_EXPR.
12374 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12376 * doc/sourcebuild.texi (vect_double_cond_arith): Include
12377 multiplication and division.
12378 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
12379 (cond_udiv@var{m}, cond_umod@var{m}): Document.
12380 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
12381 (cond_udiv_optab, cond_umod_optab): New optabs.
12382 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
12383 (IFN_COND_RDIV): New internal functions.
12384 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
12385 TRUNC_MOD_EXPR and RDIV_EXPR.
12386 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
12387 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
12389 (SVE_INT_BINARY): Include mult.
12390 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
12391 (optab, sve_int_op): Handle mult.
12392 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
12394 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
12395 for SVE_INT_BINARY_SD.
12397 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12399 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
12400 (optab, sve_int_op): Handle div and udiv.
12401 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
12402 for SVE_INT_BINARY_SD.
12403 (*<optab><mode>3): New insn for the same.
12405 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12407 * tree-vect-patterns.c: Include predict.h.
12408 (vect_recog_divmod_pattern): Restrict check for division support
12409 to when optimizing for size.
12411 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12413 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
12414 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
12415 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
12416 (gimple_match_op::set_op): Likewise.
12417 (gimple_resimplify4): Declare.
12418 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
12419 (expr::gen_transform): Likewise.
12420 (decision_tree::gen): Generate a simplification routine for 4 operands.
12421 * gimple-match-head.c (gimple_simplify): Add an overload for
12422 4 operands. In the top-level function, handle up to 4 call
12423 arguments and call gimple_resimplify4.
12424 (gimple_resimplify4): New function.
12425 (build_call_internal): Pass a fourth operand.
12426 (maybe_push_to_seq): Likewise.
12427 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
12428 Fold VEC_COND_EXPRs of an operation and a default value into
12429 an IFN_COND_* function if possible.
12430 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
12432 (SVE_COND_FP_BINARY): Include them.
12433 (optab, sve_fp_op): Handle them.
12434 (SVE_INT_BINARY_REV): New code iterator.
12435 (SVE_COND_FP_BINARY_REV): New int iterator.
12436 (commutative): New int attribute.
12437 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
12439 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
12441 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
12442 (*cond_<optab><mode>): New patterns for reversed operands.
12444 2018-05-25 Richard Biener <rguenther@suse.de>
12446 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12447 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12448 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12449 (STMT_VINFO_GROUPED_ACCESS): Adjust.
12450 * tree-vect-data-refs.c (everywhere): Adjust users.
12451 * tree-vect-loop.c (everywhere): Likewise.
12452 * tree-vect-slp.c (everywhere): Likewise.
12453 * tree-vect-stmts.c (everywhere): Likewise.
12454 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12456 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12458 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12460 (gcc_cv_as_section_exclude): ... this.
12461 Try Solaris as #exclude syntax.
12462 * configure: Regenerate.
12463 * config.in: Regenerate.
12464 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12466 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12467 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12469 * varasm.c (default_elf_asm_named_section): Don't check if
12470 HAVE_GAS_SECTION_EXCLUDE is defined.
12472 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12474 * doc/md.texi: Update the documentation of the cond_* optabs
12475 to mention the new final operand. Fix GET_MODE_NUNITS call.
12476 Describe the scalar case too.
12477 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12478 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12480 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12481 (get_conditional_internal_fn): Update comment.
12482 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12483 accumulator value as a final argument to conditional functions.
12484 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12485 a define_expand and add an "else" operand. Assert for now that
12486 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12487 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12488 (*cond_<optab><mode>): New patterns.
12489 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12490 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12491 (UNSPEC_COND_EOR): Delete.
12492 (optab): Remove associated mappings.
12493 (SVE_INT_BINARY): New code iterator.
12494 (sve_int_op): Remove int attribute and add "minus" to the code
12496 (SVE_COND_INT_OP): Delete.
12497 (SVE_COND_FP_OP): Rename to...
12498 (SVE_COND_FP_BINARY): ...this.
12500 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12502 * optabs.c (can_reuse_operands_p): New function.
12503 (maybe_legitimize_operands): Try to reuse the results for
12506 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12508 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12509 Add {q} suffix to insn mnemonic.
12511 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12513 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12514 (msp430_warn_func_return): New.
12516 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12518 * fold-const.c (tree_nonzero_bits): New function.
12519 * fold-const.h (tree_nonzero_bits): Likewise.
12520 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12521 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12523 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12527 * varasm.c (assemble_alias): Check ifunc_resolver only on
12530 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12533 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12534 when memory input operand is handled.
12536 2018-05-24 Luis Machado <luis.machado@linaro.org>
12538 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12540 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12542 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12544 * match.pd: Delay FMA folds until after vectorization.
12546 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12549 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12550 address check not strict.
12552 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12554 * gimple-match.h (gimple_match_op): New class.
12555 (mprts_hook): Replace parameters with a gimple_match_op *.
12556 (maybe_build_generic_op): Likewise.
12557 (gimple_simplified_result_is_gimple_val): Replace parameters with
12558 a const gimple_match_op *.
12559 (gimple_simplify): Replace code_helper * and tree * parameters with
12560 a gimple_match_op * parameter.
12561 (gimple_resimplify1): Replace code_helper *, tree and tree *
12562 parameters with a gimple_match_op * parameter.
12563 (gimple_resimplify2): Likewise.
12564 (gimple_resimplify3): Likewise.
12565 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12566 parameters with a gimple_match_op * parameter.
12567 * gimple-match-head.c (gimple_simplify): Change prototypes of
12568 auto-generated functions to take a gimple_match_op * instead of
12569 separate code_helper * and tree * parameters. Make the same
12570 change in the top-level overload and update calls to the
12571 gimple_resimplify routines. Update calls to the auto-generated
12572 functions and to maybe_push_res_to_seq in the publicly-facing
12573 operation-specific gimple_simplify overloads.
12574 (gimple_match_op::MAX_NUM_OPS): Define.
12575 (gimple_resimplify1): Replace rcode and ops with a single res_op
12576 parameter. Update call to gimple_simplify.
12577 (gimple_resimplify2): Likewise.
12578 (gimple_resimplify3): Likewise.
12579 (mprts_hook): Replace parameters with a gimple_match_op *.
12580 (maybe_build_generic_op): Likewise.
12581 (build_call_internal): Replace type, nargs and ops with
12582 a gimple_match_op *.
12583 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12584 with a single gimple_match_op *. Update calls to mprts_hook,
12585 build_call_internal and gimple_simplified_result_is_gimple_val.
12586 Factor out code that is common to the tree_code and combined_fn cases.
12587 * genmatch.c (expr::gen_transform): Replace tem_code and
12588 tem_ops with a gimple_match_op called tem_op. Update calls
12589 to the gimple_resimplify functions and maybe_push_res_to_seq.
12590 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12591 res_ops. Update call to the gimple_resimplify functions.
12592 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12593 (decision_tree::gen): Make the functions take a gimple_match_op *
12594 called res_op instead of separate res_code and res_ops parameters.
12595 Update call accordingly.
12596 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12597 and ops with a single res_op parameter. Update calls to
12598 maybe_build_generic_op and maybe_push_res_to_seq.
12599 (fold_stmt_1): Update calls to gimple_simplify and
12600 replace_stmt_with_simplification.
12601 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12602 and gimple_simplified_result_is_gimple_val.
12603 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12605 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12606 with a gimple_match_op *.
12607 (vn_nary_build_or_lookup): Likewise. Update call to
12608 vn_nary_build_or_lookup_1.
12609 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12610 gimple_match_op *. Update calls to the gimple_resimplify routines
12611 and to gimple_simplified_result_is_gimple_val.
12612 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12613 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12614 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12615 (visit_nary_op): Likewise.
12616 (visit_reference_op_load): Likewise.
12618 2018-05-23 Luis Machado <luis.machado@linaro.org>
12620 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12621 modifier for printing the step amount.
12623 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12626 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12629 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12631 * doc/sourcebuild.texi (Endianness): New subsubsection.
12633 2018-05-23 Luis Machado <luis.machado@linaro.org>
12635 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12636 <prefetch_dynamic_strides>: New const bool field.
12637 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12638 prefetch_dynamic_strides.
12639 (exynosm1_prefetch_tune): Likewise.
12640 (thunderxt88_prefetch_tune): Likewise.
12641 (thunderx_prefetch_tune): Likewise.
12642 (thunderx2t99_prefetch_tune): Likewise.
12643 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12645 (aarch64_override_options_internal): Update to set
12646 PARAM_PREFETCH_DYNAMIC_STRIDES.
12647 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12648 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12649 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12650 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12651 prefetch-dynamic-strides setting.
12653 2018-05-23 Luis Machado <luis.machado@linaro.org>
12655 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12656 <minimum_stride>: New const int field.
12657 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12658 minimum_stride field defaulting to -1.
12659 (exynosm1_prefetch_tune): Likewise.
12660 (thunderxt88_prefetch_tune): Likewise.
12661 (thunderx_prefetch_tune): Likewise.
12662 (thunderx2t99_prefetch_tune): Likewise.
12663 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12664 <default_opt_level>: Set to 3.
12665 (aarch64_override_options_internal): Update to set
12666 PARAM_PREFETCH_MINIMUM_STRIDE.
12667 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12668 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12669 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12670 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12671 stride is constant and is below the minimum stride threshold.
12673 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12675 * config/arm/arm-cpus.in (mode26): Delete.
12676 (armv4): Delete mode26 reference.
12677 * config/arm/arm.c (arm_configure_build_target): Delete use of
12680 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12682 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12684 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12685 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
12687 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
12688 (floatunsdidf2): Ditto.
12690 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12692 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12693 (fixuns_trunc<mode>si2_avx512f): Ditto.
12694 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12695 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12696 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12698 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12700 PR rtl-optimization/79985
12701 * df-scan.c (df_insn_refs_collect): Remove special case for
12702 global registers and asm statements.
12704 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12706 * extend.texi (Global Register Variables): Rewrite the bullet list.
12707 Note that the register is available for allocation. Note that access
12708 via inline asm must use constraints. Add note about async-signal
12709 handlers. Remove paragraph about automagic register selection.
12711 2018-05-23 Richard Biener <rguenther@suse.de>
12713 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12714 of fixed offset from memset VN.
12716 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
12718 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12719 first_interp field.
12720 (alloc_cand_and_find_basis): Initialize first_interp field.
12721 (slsr_process_mul): Modify first_interp field.
12722 (slsr_process_add): Likewise.
12723 (slsr_process_cast): Modify first_interp field for each new
12725 (slsr_process_copy): Likewise.
12726 (dump_candidate): Dump first_interp field.
12727 (replace_mult_candidate): Process all interpretations, not just
12729 (replace_rhs_if_not_dup): Likewise.
12730 (replace_one_candidate): Likewise.
12732 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
12734 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12736 (aarch64_needs_frame_chain): New function.
12737 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12739 2018-05-23 Sudakshina Das <sudi.das@arm.com>
12742 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12743 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12744 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12745 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12746 as true for strict-align.
12747 (aarch64_can_inline_p): Perform checks even when callee has no
12748 attributes to check for strict alignment.
12749 * doc/extend.texi (AArch64 Function Attributes): Document
12751 * doc/invoke.texi: (AArch64 Options): Likewise.
12753 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
12755 PR tree-optimization/85853
12756 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12757 the handling of the root of the node to...
12758 (vect_slp_analyze_node_operations_1): ...this new function,
12759 and run the whole thing with the child nodes' def types
12760 set according to their SLP node's def type.
12762 2018-05-23 Richard Biener <rguenther@suse.de>
12764 PR middle-end/85874
12765 * tree-data-ref.c (create_runtime_alias_checks): Defer
12766 and ignore overflow warnings.
12768 2018-05-23 Yury Gribov <tetra2005@gmail.com>
12770 PR tree-optimization/85822
12771 * tree-vrp.c (is_masked_range_test): Fix handling of negative
12774 2018-05-23 Richard Biener <rguenther@suse.de>
12776 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12777 memset constants via native_interpret_expr.
12779 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
12782 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12784 (cgraph_node::create_alias): Likewise.
12785 (cgraph_node::get_availability): Check ifunc_resolver instead
12786 of looking up ifunc attribute.
12787 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12788 * varasm.c (do_assemble_alias): Likewise.
12789 (assemble_alias): Likewise.
12790 (default_binds_local_p_3): Likewise.
12791 * cgraph.h (cgraph_node): Add ifunc_resolver.
12792 (cgraph_node::only_called_directly_or_aliased_p): Return false
12793 for IFUNC resolver.
12794 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12796 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12797 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12798 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12799 instead of looking up ifunc attribute.
12801 2018-05-22 Luis Machado <luis.machado@linaro.org>
12803 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12805 2018-05-22 Martin Sebor <msebor@redhat.com>
12807 PR middle-end/85359
12808 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12809 only when expasion succeeds.
12810 (expand_builtin_strcmp): Same.
12811 (expand_builtin_strncmp): Same.
12813 2018-05-22 Martin Sebor <msebor@redhat.com>
12815 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12817 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12818 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12820 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12821 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12823 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12824 Add check for legitimate_address.
12825 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12826 (aarch64_swap_ldrstr_operands): New.
12827 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12830 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12831 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12833 * config/aarch64/aarch64.md: New patterns to generate stp
12835 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12836 single words and double words.
12837 (load_pair_sw, load_pair_dw): Likewise.
12838 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12840 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12842 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12843 for different mode ldpstp and add peephole for merged zero stores.
12844 Likewise for loads.
12845 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12847 (aarch64_gen_store_pair): Rename calls to match new patterns.
12848 (aarch64_gen_load_pair): Rename calls to match new patterns.
12849 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12850 (load_pair<DREG:mode><DREG2:mode>): ... This.
12851 (store_pair<mode>): Rename to...
12852 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12853 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12854 New mode iterators.
12855 (V_INT_EQUIV): Handle SImode.
12856 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12859 2018-05-22 Martin Sebor <msebor@redhat.com>
12862 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12863 or ajust the presumed bound on an operation to avoid unnecessary
12866 2018-05-22 Martin Sebor <msebor@redhat.com>
12868 PR tree-optimization/85826
12869 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12870 assuming that a DECL necesarily has a constant size.
12872 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12874 PR middle-end/85862
12875 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12877 2018-05-22 Richard Biener <rguenther@suse.de>
12879 PR tree-optimization/85834
12880 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12881 non-constant and non-zero memset arguments.
12883 2018-05-22 Martin Liska <mliska@suse.cz>
12886 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12888 2018-05-22 Richard Biener <rguenther@suse.de>
12890 PR tree-optimization/85863
12891 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12892 comparisons when vectype is specified.
12893 (vectorizable_condition): Do not specify vectype for
12894 vect_is_simple_cond when SLP vectorizing.
12896 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12899 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12900 define __ibm128 as long double.
12901 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12902 as a distinct type when IEEE 128-bit support is enabled.
12903 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12904 128-bit types to use the correct functions.
12905 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12906 convert between 128-bit floating point types that have different
12907 modes but the same representation, instead of using gen_lowpart to
12909 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12911 (IFKF_reg): New attributes to give the register constraints for
12913 (extend<mode>tf2_internal): New insns to mark an explicit
12914 conversion between 128-bit floating point types that have a
12915 different mode but share the same representation.
12917 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12919 PR tree-optimization/85814
12920 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12921 a null return from get_strinfo when unsharing the next
12922 strinfo in the chain.
12924 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12927 * varasm.c (weak_finish): Clean up weak_decls.
12929 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12931 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12932 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12933 UNSPEC_UADALP values.
12934 * config/aarch64/iterators.md (ABAL): New int iterator.
12937 (sur): Add mappings for the above.
12938 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12940 (aarch64_<sur>abal<mode>_4): Likewise.
12941 (aarch64_<sur>adalp<mode>_3): Likewise.
12942 (<sur>sadv16qi): New define_expand.
12944 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12946 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12947 (*movdf_internal): Ditto.
12948 (*rcpsf2_sse): Ditto.
12949 (*rsqrtsf2_sse): Ditto.
12950 (*sqrt<mode>2_sse): Ditto.
12952 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12954 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12956 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12957 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12958 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12959 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12961 * config/aarch64/arm_neon.h: Likewise.
12962 * config/aarch64/iterators.md (VQ_I): New.
12964 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12966 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12968 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12971 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12973 * config/nds32/constraints.md (S): New constraint.
12974 * config/nds32/nds32.md (call_internal): Use constraint S.
12975 (call_value_internal): Likewise.
12976 (sibcall_internal): Likewise.
12977 (sibcall_value_internal): Likewise.
12979 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12980 Chung-Ju Wu <jasonwucj@gmail.com>
12982 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12983 into consideration.
12985 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12986 Chung-Ju Wu <jasonwucj@gmail.com>
12988 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12989 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12990 (nds32_rtx_costs_impl): Simplify.
12991 (nds32_address_cost_impl): Simplify.
12992 (nds32_init_rtx_costs): New function.
12993 (nds32_rtx_costs_speed_prefer): Likewise.
12994 (nds32_rtx_costs_size_prefer): Likewise.
12995 (nds32_address_cost_speed_prefer): Likewise.
12996 (nds32_address_cost_speed_fwprop): Likewise.
12997 (nds32_address_cost_size_prefer): Likewise.
12998 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12999 * config/nds32/nds32.c (nds32_option_override): Use
13000 nds32_init_rtx_costs function.
13002 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13004 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
13005 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
13006 (TARGET_PIPELINE_N8): Likewise.
13007 (TARGET_PIPELINE_N10): Likewise.
13008 (TARGET_PIPELINE_N13): Likewise.
13009 (TARGET_PIPELINE_GRAYWOLF): Likewise.
13011 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
13013 * config/nds32/nds32-fpu.md: Update copyright year.
13015 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13017 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
13019 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13021 * config/nds32/nds32.c
13022 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13023 * config/nds32/nds32.opt (minline-asm-r15): New option.
13025 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13027 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
13029 * config/nds32/nds32.md (abssi2): New pattern.
13031 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
13033 * config/i386/i386.md (rex64namesuffix): New mode attribute.
13034 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
13035 Merge insn pattern from sse_cvtsi2ss<round_name> and
13036 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13037 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
13038 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
13039 using SWI48 mode iterator.
13040 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
13041 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
13042 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13043 pattern from sse_cvttss2si<round_saeonly_name>
13044 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
13045 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
13046 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
13047 using SWI48 mode iterator.
13048 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
13049 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
13050 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
13051 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
13052 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
13053 using SWI48 mode iterator.
13054 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
13055 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
13056 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
13057 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
13058 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
13059 SWI48 mode iterator.
13060 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
13061 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
13062 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
13063 pattern from sse_cvttsd2si<round_saeonly_name>
13064 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
13066 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13068 * config/nds32/nds32-md-auxiliary.c
13069 (nds32_valid_smw_lwm_base_p): Refine.
13070 (nds32_output_smw_single_word): Refine.
13071 (nds32_output_smw_double_word): New.
13072 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
13074 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13076 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
13077 (nds32_output_stack_pop): Refine.
13078 (nds32_expand_unaligned_load): Refine.
13079 (nds32_expand_unaligned_store): Refine.
13081 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13082 Chung-Ju Wu <jasonwucj@gmail.com>
13084 * config/nds32/constants.md: Add TP_REGNUM constant.
13085 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
13086 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
13088 * config/nds32/nds32-doubleword.md: Consider flag_pic.
13089 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
13090 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
13091 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
13092 and PIC code generation.
13093 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
13095 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
13097 * config/nds32/nds32.md: Support TLS and PIC.
13098 * config/nds32/nds32.c: Support TLS and PIC.
13099 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
13100 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
13103 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13105 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
13106 mode with E_ prefix.
13108 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13109 Chung-Ju Wu <jasonwucj@gmail.com>
13111 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
13112 * config/nds32/nds32-md-auxiliary.c
13113 (symbolic_reference_mentioned_p): New.
13114 (nds32_legitimize_ict_address): New.
13115 (nds32_expand_ict_move): New.
13116 (nds32_indirect_call_referenced_p): New.
13117 (nds32_symbol_binds_local_p): Delete.
13118 (nds32_long_call_p): Modify.
13119 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13120 * config/nds32/nds32-protos.h
13121 (symbolic_reference_mentioned_p): Declare.
13122 (nds32_legitimize_ict_address): Declare.
13123 (nds32_expand_ict_move): Declare.
13124 (nds32_indirect_call_referenced_p): Declare.
13125 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
13126 (nds32_relax_group): Use nds32_ict_const_p as condition.
13127 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
13128 (nds32_asm_file_start): Output ict_model directive in asm code.
13129 (nds32_legitimate_address_p): Consider indirect call.
13130 (nds32_print_operand): Consider indirect call.
13131 (nds32_print_operand_address): Consider indirect call.
13132 (nds32_insert_attributes): Handle "indirect_call" attribute.
13133 (TARGET_LEGITIMATE_ADDRESS_P): Define.
13134 (TARGET_LEGITIMATE_CONSTANT_P): Define.
13135 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
13136 (TARGET_DELEGITIMIZE_ADDRESS): Define.
13137 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
13138 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
13139 (TARGET_ICT_MODEL_SMALL): Define.
13140 (TARGET_ICT_MODEL_LARGE): Define.
13141 * config/nds32/nds32.md (movsi): Consider ict model.
13142 (call, call_value): Consider ict model.
13143 (sibcall, sibcall_value): Consider ict model.
13144 * config/nds32/nds32.opt (mict-model): New option.
13145 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
13148 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
13149 Monk Chiang <sh.chiang04@gmail.com>
13150 Jim Wilson <jimw@sifive.com>
13152 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
13153 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
13154 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
13155 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
13156 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
13157 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
13158 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
13159 compute save_libcall_adjustment properly.
13160 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
13161 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
13162 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
13163 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
13164 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
13165 (ABI_SPEC): Handle mabi=ilp32e.
13166 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
13167 (RVE): Add RVE mask.
13168 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
13169 <-march>: Add rv32e as an example.
13171 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
13174 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
13175 (intra_create_variable_infos): Handle C++ constructors.
13177 2018-05-18 Martin Liska <mliska@suse.cz>
13179 * passes.def: Remove a redundant pass.
13181 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
13184 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
13186 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13188 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
13190 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
13192 (armv2, armv2a, armv3, armv3m): Delete architectures.
13193 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
13194 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
13195 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
13197 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13198 (*mulsidi3adddi): Likewise.
13199 (mulsidi3): Likewise.
13200 (*mulsidi3_nov6): Likewise.
13201 (umulsidi3): Likewise.
13202 (umulsidi3_nov6): Likewise.
13203 (umaddsidi4): Likewise.
13204 (*umulsidi3adddi): Likewise.
13205 (smulsi3_highpart): Likewise.
13206 (*smulsi3_highpart_nov6): Likewise.
13207 (umulsi3_highpart): Likewise.
13208 (*umulsi3_highpart_nov6): Likewise.
13209 * config/arm/arm.h (arm_arch3m): Delete.
13210 * config/arm/arm.c (arm_arch3m): Delete.
13211 (arm_option_override_internal): Update armv3-related comment.
13212 (arm_configure_build_target): Delete use of isa_bit_mode32.
13213 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
13214 (arm_rtx_costs_internal): Delete check of arm_arch3m.
13215 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
13216 (mulsa3): Likewise.
13217 (mulusa3): Likewise.
13218 * config/arm/arm-protos.h (arm_arch3m): Delete.
13219 * config/arm/arm-tables.opt: Regenerate.
13220 * config/arm/arm-tune.md: Likewise.
13221 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
13222 deleted architectures.
13224 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13226 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
13227 (armv5t, armv5te): New features.
13228 (ARMv5, ARMv5e): Delete fgroups.
13229 (ARMv5t, ARMv5te): Adjust for above changes.
13230 (ARMv6m): Likewise.
13231 (armv5, armv5e): Delete arches.
13232 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
13234 (*call_reg_arm): Likewise.
13235 (*call_value_reg_armv5): Likewise.
13236 (*call_value_reg_arm): Likewise.
13237 (*call_symbol): Likewise.
13238 (*call_value_symbol): Likewise.
13239 (*sibcall_insn): Likewise.
13240 (*sibcall_value_insn): Likewise.
13241 (clzsi2): Likewise.
13242 (prefetch): Likewise.
13243 (define_split and define_peephole2 dependent on arm_arch5):
13245 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
13247 (TARGET_ARM_QBIT): Likewise.
13248 (TARGET_DSP_MULTIPLY): Likewise.
13249 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
13250 (arm_arch5, arm_arch5e): Delete.
13251 (arm_arch5t, arm_arch5te): Declare.
13252 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
13253 (arm_arch5t): Declare.
13254 (arm_option_reconfigure_globals): Update for the above.
13255 (arm_options_perform_arch_sanity_checks): Update comment, replace
13256 use of arm_arch5 with arm_arch5t.
13257 (use_return_insn): Likewise.
13258 (arm_emit_call_insn): Likewise.
13259 (output_return_instruction): Likewise.
13260 (arm_final_prescan_insn): Likewise.
13261 (arm_coproc_builtin_available): Likewise.
13262 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
13263 arm_arch5e with arm_arch5t and arm_arch5te.
13264 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
13265 (arm_arch5t, arm_arch5te): Declare.
13266 * config/arm/arm-tables.opt: Regenerate.
13267 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
13268 * config/arm/t-multilib: Likewise.
13269 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
13270 instead of arm_arch5.
13271 (*call_reg_thumb1): Likewise.
13272 (*call_value_reg_thumb1_v5): Likewise.
13273 (*call_value_reg_thumb1): Likewise.
13274 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
13276 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
13278 2018-05-18 Martin Liska <mliska@suse.cz>
13280 PR gcov-profile/84846
13281 * doc/gcov.texi: Document -t option of gcov tool.
13283 2018-05-18 Martin Liska <mliska@suse.cz>
13285 PR gcov-profile/84846
13286 * gcov.c (print_usage): Add new -t option.
13287 (process_args): Handle the option.
13288 (generate_results): Use stdout as output when requested by
13291 2018-05-18 Martin Liska <mliska@suse.cz>
13293 PR gcov-profile/84846
13294 * coverage.c (coverage_init): Write PWD to .gcno file.
13295 * doc/gcov.texi: Document how working directory is printed.
13296 * gcov-dump.c (dump_gcov_file): Print PWD.
13297 * gcov.c (output_intermediate_file): Likewise.
13298 (read_graph_file): Read PWD string.
13299 (output_lines): Print PWD.
13301 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13303 PR middle-end/85817
13304 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
13305 for retval and return false if all args to phi are zero.
13307 2018-05-18 Richard Biener <rguenther@suse.de>
13309 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
13311 (evrp_dom_walker::before_dom_children): Call it.
13313 2018-05-18 Richard Biener <rguenther@suse.de>
13315 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
13316 results when processing array refs with variable index.
13318 2018-05-18 Toon Moene <toon@moene.org>
13320 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
13321 directly after that of -floop-interchange. Indicate that both
13322 options are enabled by default when specifying -O3.
13324 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13326 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
13327 iterator. Delete separate integer-mode vec_set<mode> expander.
13328 (aarch64_simd_vec_setv2di): Delete.
13329 (vec_setv2di): Delete.
13330 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
13331 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
13332 the "w, r" alternative.
13334 2018-05-18 Martin Liska <mliska@suse.cz>
13336 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
13337 * tree-pass.h (make_pass_lower_switch_O0): New function.
13338 * tree-switch-conversion.c (node_has_low_bound): Remove.
13339 (node_has_high_bound): Likewise.
13340 (node_is_bounded): Likewise.
13341 (class pass_lower_switch): Make it a template type and create
13343 (pass_lower_switch::execute): Add template argument.
13344 (make_pass_lower_switch): New function.
13345 (make_pass_lower_switch_O0): New function.
13346 (do_jump_if_equal): Remove.
13347 (emit_case_nodes): Simplify to just handle all 3 cases and leave
13348 all the hard work to tree optimization passes.
13350 2018-05-18 Martin Liska <mliska@suse.cz>
13352 * dbgcnt.c (limit_low): Renamed from limit.
13353 (limit_high): New variable.
13354 (dbg_cnt_is_enabled): Check for upper limit.
13355 (dbg_cnt): Adjust dumping.
13356 (dbg_cnt_set_limit_by_index): Add new argument for high
13358 (dbg_cnt_set_limit_by_name): Likewise.
13359 (dbg_cnt_process_single_pair): Parse new format.
13360 (dbg_cnt_process_opt): Use strtok.
13361 (dbg_cnt_list_all_counters): Remove 'value' and add
13363 * doc/invoke.texi: Document changes.
13365 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
13367 * doc/sourcebuild.texi (scalar_all_fma): Document.
13368 * tree.def (FMA_EXPR): Delete.
13369 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
13370 * internal-fn.c (ternary_direct): New macro.
13371 (expand_ternary_optab_fn): Likewise.
13372 (direct_ternary_optab_supported_p): Likewise.
13373 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
13374 * builtins.c (fold_builtin_fma): Delete.
13375 (fold_builtin_3): Don't call it.
13376 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
13377 * expr.c (expand_expr_real_2): Likewise.
13378 * fold-const.c (operand_equal_p): Likewise.
13379 (fold_ternary_loc): Likewise.
13380 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
13381 * gimple.c (DEFTREECODE): Likewise.
13382 * gimplify.c (gimplify_expr): Likewise.
13383 * optabs-tree.c (optab_for_tree_code): Likewise.
13384 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
13385 * tree-eh.c (operation_could_trap_p): Likewise.
13386 (stmt_could_throw_1_p): Likewise.
13387 * tree-inline.c (estimate_operator_cost): Likewise.
13388 * tree-pretty-print.c (dump_generic_node): Likewise.
13389 (op_code_prio): Likewise.
13390 * tree-ssa-loop-im.c (stmt_cost): Likewise.
13391 * tree-ssa-operands.c (get_expr_operands): Likewise.
13392 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
13393 * fold-const-call.h (fold_fma): Delete.
13394 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
13395 CFN_FNMA and CFN_FNMS.
13396 (fold_fma): Delete.
13397 * genmatch.c (combined_fn): New enum.
13398 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
13399 (commutative_op): New function.
13400 (commutate): Use it. Handle more than 2 operands.
13401 (dt_operand::gen_gimple_expr): Use commutative_op.
13402 (parser::parse_expr): Allow :c to be used with non-binary
13403 operators if the commutative operand is known.
13404 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
13405 CFN_FMS, CFN_FNMA and CFN_FNMS.
13406 (backprop::process_assign_use): Remove FMA_EXPR handling.
13407 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
13408 (gen_hsa_fma): New function.
13409 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
13410 IFN_FNMA and IFN_FNMS.
13411 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
13412 * gimple-fold.h (follow_all_ssa_edges): Declare.
13413 * gimple-fold.c (follow_all_ssa_edges): New function.
13414 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
13415 gimple_build interface and use follow_all_ssa_edges to fold the result.
13416 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
13417 instead of checking for optabs directly.
13418 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
13419 rather than FMA_EXPRs.
13420 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
13421 call to IFN_FMA instead of an FMA_EXPR.
13423 2018-05-17 Jim Wilson <jimw@sifive.com>
13425 * expr.c (do_tablejump): When converting index to Pmode, if we have a
13426 sign extended promoted subreg, and the range does not have the sign bit
13427 set, then do a sign extend.
13429 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
13430 test, check for sign extended subreg and/or constant operands, and
13431 do a sign extend in that case.
13433 2018-05-17 Steve Ellcey <sellcey@cavium.com>
13435 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
13436 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
13438 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
13439 Change logics_shift_reg to logics_shift_imm.
13440 (thunderx2t99_fp_loadpair_basic): Delete.
13441 (thunderx2t99_fp_storepair_basic): Delete.
13442 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13443 (thunderx2t99_asimd_polynomial): Delete.
13444 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13445 and neon_fp_mul_d_scalar_q.
13446 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13447 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13448 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13449 (thunderx2t99_asimd_lut): Add missing tbl types.
13450 (thunderx2t99_asimd_ext): Delete.
13451 (thunderx2t99_asimd_load1_1_mult): Delete.
13452 (thunderx2t99_asimd_load1_2_mult): Delete.
13453 (thunderx2t99_asimd_load1_ldp): New.
13454 (thunderx2t99_asimd_load1): New.
13455 (thunderx2t99_asimd_load2): Add missing *load2* types.
13456 (thunderx2t99_asimd_load3): New.
13457 (thunderx2t99_asimd_load4): New.
13458 (thunderx2t99_asimd_store1_1_mult): Delete.
13459 (thunderx2t99_asimd_store1_2_mult): Delete.
13460 (thunderx2t99_asimd_store2_mult): Delete.
13461 (thunderx2t99_asimd_store2_onelane): Delete.
13462 (thunderx2t99_asimd_store_stp): New.
13463 (thunderx2t99_asimd_store1): New.
13464 (thunderx2t99_asimd_store2): New.
13465 (thunderx2t99_asimd_store3): New.
13466 (thunderx2t99_asimd_store4): New.
13468 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
13470 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13471 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
13473 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
13474 Segher Boessenkool <segher@kernel.crashing.org>
13477 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13480 2018-05-17 Richard Biener <rguenther@suse.de>
13482 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13483 for pruning loop and prune defs feeding only already visited PHIs.
13485 2018-05-17 Richard Biener <rguenther@suse.de>
13487 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13489 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13490 Richard Biener <rguenther@suse.de>
13492 PR tree-optimization/85793
13493 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13494 for VMAT_ELEMENTWISE.
13496 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13498 * internal-fn.h (lookup_internal_fn): Declare
13499 * internal-fn.c (lookup_internal_fn): New function.
13500 * gimple.c (gimple_build_call_from_tree): Handle calls to
13501 internal functions.
13502 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13503 internal function names.
13504 * tree-pretty-print.c (dump_generic_node): Likewise.
13505 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13507 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13509 * gimple-fold.h (gimple_build): Make the function forms take
13510 combined_fn rather than built_in_function.
13511 (gimple_simplify): Likewise.
13512 * gimple-match-head.c (gimple_simplify): Likewise.
13513 * gimple-fold.c (gimple_build): Likewise.
13514 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13515 rather than gimple_build_call_internal.
13516 (get_initial_defs_for_reduction): Likewise.
13517 (vect_create_epilog_for_reduction): Likewise.
13518 (vectorizable_live_operation): Likewise.
13520 2018-05-17 Martin Liska <mliska@suse.cz>
13522 * gimple-ssa-sprintf.c (format_directive): Do not use
13523 space in between 'G_' and '('.
13525 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13528 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13529 even if the mask is not all ones.
13532 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13534 (ix86_gimple_fold_builtin): Likewise.
13537 * config/i386/i386.c: Include tree-vector-builder.h.
13538 (ix86_vector_shift_count): New function.
13539 (ix86_fold_builtin): Fold shift builtins by scalar count.
13540 (ix86_gimple_fold_builtin): Likewise.
13542 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13543 _mm512_setzero): New intrinsics.
13545 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13546 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13548 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13549 code generation for cases where splatting a value is not useful.
13550 * simplify-rtx.c (simplify_ternary_operation): Simplify
13551 vec_merge across a vec_duplicate and a paradoxical subreg forming
13552 a vector mode to a vec_concat.
13554 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13556 * config.gcc: Support "goldmont-plus".
13557 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13559 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13560 PROCESSOR_GOLDMONT_PLUS.
13561 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13562 (processor_target_table): Add "goldmont-plus".
13563 (PTA_GOLDMONT_PLUS): Define.
13564 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13565 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13566 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13567 (fold_builtin_cpu): Add "goldmont-plus".
13568 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13569 (ix86_option_override_internal): Add "goldmont-plus".
13570 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13571 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13572 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13573 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13575 2018-05-17 Richard Biener <rguenther@suse.de>
13577 PR tree-optimization/85757
13578 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13579 remove defs that only feed that PHI from further processing.
13581 2018-05-16 Jim Wilson <jimw@sifive.com>
13583 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13585 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13587 2018-05-16 Mark Wielaard <mark@klomp.org>
13589 * dwarf2out.c (count_index_strings): New function.
13590 (output_indirect_strings): Call count_index_strings and generate
13591 header for dwarf_version >= 5.
13593 2018-05-16 Mark Wielaard <mark@klomp.org>
13595 * dwarf2out.c (dwarf_FORM): New function.
13596 (set_indirect_string): Use dwarf_FORM.
13597 (reset_indirect_string): Likewise.
13598 (size_of_die): Likewise.
13599 (value_format): Likewise.
13600 (output_die): Likewise.
13601 (add_skeleton_AT_string): Likewise.
13602 (output_macinfo_op): Likewise.
13603 (index_string): Likewise.
13604 (output_index_string_offset): Likewise.
13605 (output_index_string): Likewise.
13606 (count_index_strings): Likewise.
13608 2018-05-16 Carl Love <cel@us.ibm.com>
13610 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13611 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13613 2018-05-16 Martin Jambor <mjambor@suse.cz>
13615 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13616 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13618 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13620 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13621 (fnma<mode>4): Likewise.
13622 (fms<mode>4): Likewise.
13623 (fnms<mode>4): Likewise.
13624 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13625 (aarch64_fnma<mode>4): Likewise.
13626 (aarch64_fms<mode>4): Likewise.
13627 (aarch64_fnms<mode>4): Likewise.
13628 (aarch64_fnmadd<mode>4): Likewise.
13630 2018-05-16 Jason Merrill <jason@redhat.com>
13632 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13634 2018-05-16 Richard Biener <rguenther@suse.de>
13636 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13637 (dump_stmt_cost): Declare.
13638 (add_stmt_cost): Dump cost we add.
13639 (add_stmt_costs): New function.
13640 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13641 No longer exported.
13642 (vect_analyze_stmt): Adjust prototype.
13643 (vectorizable_condition): Likewise.
13644 (vectorizable_live_operation): Likewise.
13645 (vectorizable_reduction): Likewise.
13646 (vectorizable_induction): Likewise.
13647 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13648 cost vector to pass to vectorizable_ and record afterwards.
13649 (vect_model_reduction_cost): Take cost vector argument and adjust.
13650 (vect_model_induction_cost): Likewise.
13651 (vectorizable_reduction): Likewise.
13652 (vectorizable_induction): Likewise.
13653 (vectorizable_live_operation): Likewise.
13654 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13655 SLP_TREE_NUMBER_OF_VEC_STMTS.
13656 (vect_analyze_slp_cost_1): Remove.
13657 (vect_analyze_slp_cost): Likewise.
13658 (vect_slp_analyze_node_operations): Take visited args and
13659 a target cost vector. Avoid processing already visited stmt sets.
13660 (vect_slp_analyze_operations): Use a local cost vector to gather
13661 costs and register those of non-discarded instances.
13662 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13663 (vect_schedule_slp_instance): Remove copying of
13664 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13666 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13667 adding cost. Record cost entry location.
13668 (vect_prologue_cost_for_slp_op): Function to compute cost of
13669 a constant or invariant generated for SLP vect in the prologue,
13670 split out from vect_analyze_slp_cost_1.
13671 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13672 (vect_model_promotion_demotion_cost): Likewise.
13673 (vect_model_store_cost): Likewise, make static.
13674 (vect_model_load_cost): Likewise.
13675 (vectorizable_bswap): Add cost vector arg and adjust.
13676 (vectorizable_call): Likewise.
13677 (vectorizable_simd_clone_call): Likewise.
13678 (vectorizable_conversion): Likewise.
13679 (vectorizable_assignment): Likewise.
13680 (vectorizable_shift): Likewise.
13681 (vectorizable_operation): Likewise.
13682 (vectorizable_store): Likewise.
13683 (vectorizable_load): Likewise.
13684 (vectorizable_condition): Likewise.
13685 (vectorizable_comparison): Likewise.
13686 (can_vectorize_live_stmts): Likewise.
13687 (vect_analyze_stmt): Likewise.
13688 (vect_transform_stmt): Adjust calls to vectorizable_*.
13689 * tree-vectorizer.c: Include gimple-pretty-print.h.
13690 (dump_stmt_cost): New function.
13692 2018-05-16 Richard Biener <rguenther@suse.de>
13694 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13695 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13696 * tree-ssa-dse.c: Include tree-ssa-loop.h.
13697 (check_name): New callback.
13698 (dse_classify_store): Track cycles via a visited bitmap of PHI
13699 defs and simplify handling of in-loop and across loop dead stores
13700 and properly fail for loop-variant refs. Handle byte-tracking with
13701 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13704 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
13706 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13707 (vect_get_mask_type_for_stmt): Likewise.
13708 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13710 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
13711 to determine the statement's vector type and the vector type that
13712 should be used for calculating nunits. Deal with cases in which
13713 the type has to be deferred.
13714 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13715 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13716 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13717 (vect_determine_vf_for_stmt): New functions, split out from...
13718 (vect_determine_vectorization_factor): ...here.
13719 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13720 (vect_get_mask_type_for_stmt): New functions, split out from
13721 vect_determine_vectorization_factor.
13723 2018-05-16 Richard Biener <rguenther@suse.de>
13725 * tree-cfg.c (verify_gimple_assign_ternary): Properly
13726 verify the [VEC_]COND_EXPR embedded comparison.
13728 2018-05-15 Martin Sebor <msebor@redhat.com>
13730 PR tree-optimization/85753
13731 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13732 RECORD_TYPE in addition to ARRAY_TYPE.
13734 2018-05-15 Martin Sebor <msebor@redhat.com>
13736 PR middle-end/85643
13737 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13739 2018-05-15 Richard Biener <rguenther@suse.de>
13741 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13742 add by_clobber_p one. Change algorithm to collect all defs
13743 representing uses we need to walk and try reducing them to
13744 a single one before failing.
13745 (dse_dom_walker::dse_optimize_stmt): Adjust.
13747 2018-05-13 Mark Wielaard <mark@klomp.org>
13749 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13750 (size_of_loc_descr): Likewise.
13751 (output_loc_operands): Likewise.
13752 (output_loc_operands_raw): Likewise.
13753 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13754 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13755 (hash_loc_operands): Likewise.
13756 (compare_loc_operands): Likewise.
13758 2018-05-14 Mark Wielaard <mark@klomp.org>
13760 * dwarf2out.c (count_index_addrs): New function.
13761 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13763 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13765 PR tree-optimization/83648
13766 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13767 return value as malloc candidate.
13769 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13772 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13773 param as true in call to suggest_attribute.
13775 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
13777 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13778 -mreadonly-in-sdata.
13780 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13782 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13784 (aarch64_crypto_aesd_fused): Likewise.
13786 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
13788 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13789 (movsi_aarch64): Likewise.
13790 (load_pairsi): Likewise.
13791 (load_pairdi): Likewise.
13792 (store_pairsi): Likewise.
13793 (store_pairdi): Likewise.
13794 (load_pairsf): Likewise.
13795 (load_pairdf): Likewise.
13796 (store_pairsf): Likewise.
13797 (store_pairdf): Likewise.
13798 (zero_extend): Likewise.
13799 (trunc): Swap alternatives.
13800 (fcvt_target): Add '?' to prefer w over r.
13802 2018-05-14 Jakub Jelinek <jakub@redhat.com>
13805 * config/i386/i386.md: Disallow non-commutative arithmetics in
13806 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13807 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
13808 in the peephole2 before it.
13810 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13812 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13813 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13814 (ix86_handle_option): Handle -mcldemote.
13815 * config.gcc: New header.
13816 * config/i386/cldemoteintrin.h: New file.
13817 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13818 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13820 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13821 OPTION_MASK_ISA_CLDEMOTE.
13822 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13823 (ix86_valid_target_attribute_inner_p): Ditto.
13824 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13825 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13826 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13827 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13828 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13830 * config/i386/i386.opt: Add -mcldemote.
13831 * config/i386/x86intrin.h: New header.
13832 * doc/invoke.texi: Add -mcldemote.
13834 2018-05-14 Richard Biener <rguenther@suse.de>
13836 * doc/match-and-simplify.texi: Adjust :s documentation.
13838 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13840 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13841 intended memcpy size.
13842 (REORDER_45): Likewise.
13844 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13846 * sort.cc: New file.
13847 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13848 * vec.c (qsort_chk): Use gcc_qsort.
13849 * Makefile.in (OBJS-libcommon): Add sort.o.
13850 (build/sort.o): New target. Use it...
13851 (BUILD_RTL): ... here, and...
13852 (build/gencfn-macros): ... here, and...
13853 (build/genmatch): ... here.
13855 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13856 Chung-Ju Wu <jasonwucj@gmail.com>
13858 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13859 * config/nds32/nds32-graywolf.md: New file.
13860 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13861 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13863 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13864 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13865 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13866 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13867 * config/nds32/pipelines.md: Include n15 settings.
13869 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13870 Chung-Ju Wu <jasonwucj@gmail.com>
13872 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13873 * config/nds32/nds32-n13.md: New file.
13874 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13875 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13877 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13878 * config/nds32/nds32.md (pipeline_model): Add n13.
13879 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13880 * config/nds32/pipelines.md: Include n13 settings.
13882 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13883 Chung-Ju Wu <jasonwucj@gmail.com>
13885 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13886 * config/nds32/nds32-n10.md: New file.
13887 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13888 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13890 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13891 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13892 * config/nds32/nds32.md (pipeline_model): Add n10.
13893 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13894 * config/nds32/pipelines.md: Include n10 settings.
13896 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13897 Kito Cheng <kito.cheng@gmail.com>
13898 Chung-Ju Wu <jasonwucj@gmail.com>
13900 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13901 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13902 Add enum values for DSP extension instructions.
13903 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13905 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13906 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13907 New code iterators.
13908 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13909 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13910 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13911 * config/nds32/nds32-intrinsic.md: Likewise.
13912 * config/nds32/nds32_intrinsic.h: Likewise.
13913 * config/nds32/nds32-md-auxiliary.c: Likewise.
13914 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13915 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13916 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13917 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13918 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13919 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13920 TYPE_DMAC in switch statement.
13921 * config/nds32/nds32.c: New checking and implementation for DSP
13922 extension instructions.
13923 * config/nds32/nds32.h: Likewise.
13924 * config/nds32/nds32.md: Likewise.
13925 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13926 * config/nds32/predicates.md: Implement new predicates for DSP
13929 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13931 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13932 Reformat alternatives and attributes so it is easier to identify
13933 which constraints/attributes go with which instruction.
13934 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13935 (mov<mode>_softfloat32, FMOVE64): Likewise.
13936 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13937 (mov<mode>_softfloat64, FMOVE64): Likewise.
13939 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13941 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13943 (Basic PowerPC Built-in Functions): The new name of the
13944 subsection previously known as "PowerPC Built-in Functions".
13945 (Basic PowerPC Built-in Functions Available on all Configurations):
13947 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13948 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13949 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13950 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13952 2018-05-11 Martin Jambor <mjambor@suse.cz>
13955 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13958 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13961 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13963 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13965 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13966 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13967 (ix86_handle_option): Handle -mwaitpkg.
13968 * config.gcc: New header.
13969 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13970 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13971 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13973 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13974 OPTION_MASK_ISA_WAITPKG.
13975 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13976 (ix86_option_override_internal): Add PTA_WAITPKG.
13977 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13978 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13979 IX86_BUILTIN_TPAUSE.
13980 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13981 __builtin_ia32_umwait and __builtin_ia32_tpause.
13982 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13983 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13984 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13985 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13986 UNSPECV_TPAUSE): New.
13987 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13988 * config/i386/i386.opt: Add -mwaitpkg.
13989 * config/i386/waitpkgintrin.h: New file.
13990 * config/i386/x86intrin.h: New header.
13991 * doc/invoke.texi: Add -mwaitpkg.
13993 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13996 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13998 (cortex-m0): Use armv6s-m isa.
13999 (cortex-m0plus): Likewise.
14000 (cortex-m1): Likewise.
14001 (cortex-m0.small-multiply): Likewise.
14002 (cortex-m0plus.small-multiply): Likewise.
14003 (cortex-m1.small-multiply): Likewise.
14005 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
14006 Jakub Jelinek <jakub@redhat.com>
14008 PR tree-optimization/85692
14009 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
14010 source permute as well.
14012 2018-05-11 Martin Liska <mliska@suse.cz>
14015 * doc/extend.texi: Document LLVM style format for no_sanitize
14018 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14020 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
14021 mode_supports_vsx_dform_quad to mode_supports_dq_form.
14022 (mode_supports_vsx_dform_quad): Likewise.
14023 (mode_supports_vmx_dform): Move these functions to be next to the
14024 other mode_supports functions.
14025 (mode_supports_dq_form): Likewise.
14026 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
14027 mode_supports_dq_form.
14028 (reg_offset_addressing_ok_p): Likewise.
14029 (offsettable_ok_by_alignment): Likewise.
14030 (rs6000_legitimate_offset_address_p): Likewise.
14031 (legitimate_lo_sum_address_p): Likewise.
14032 (rs6000_legitimize_address): Likewise.
14033 (rs6000_legitimize_reload_address): Likewise.
14034 (rs6000_secondary_reload_inner): Likewise.
14035 (rs6000_preferred_reload_class): Likewise.
14036 (rs6000_output_move_128bit): Likewise.
14038 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14040 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
14041 Generate SImode target register for null target.
14042 <case IX86_BUILTIN_XGETBV>: Ditto.
14043 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
14044 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
14046 2018-05-10 Carl Love <cel@us.ibm.com>
14048 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14049 dcbtt and dcbtstt if operands[2] is 0.
14051 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14054 * config/i386/sse.md (usadv64qi): New expander.
14056 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
14058 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
14059 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
14060 -maltivec=be support.
14061 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
14062 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
14063 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
14064 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
14065 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
14066 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
14067 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
14068 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
14069 altivec_vsumsws): Adjust.
14070 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
14071 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
14072 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
14074 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
14075 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
14076 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
14077 (altivec_lve<VI_char>x): Delete expand.
14078 (*altivec_lve<VI_char>x_internal): Rename to...
14079 (altivec_lve<VI_char>x): ... this.
14080 (altivec_lvxl_<mode>): Delete expand.
14081 (*altivec_lvxl_<mode>_internal): Rename to ...
14082 (altivec_lvxl_<mode>): ... this.
14083 (altivec_stvxl_<mode>): Delete expand.
14084 (*altivec_stvxl_<mode>_internal): Rename to ...
14085 (altivec_stvxl_<mode>): ... this.
14086 (altivec_stve<VI_char>x): Delete expand.
14087 (*altivec_stve<VI_char>x_internal): Rename to ...
14088 (altivec_stve<VI_char>x): ... this.
14089 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
14090 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
14091 reduc_plus_scal_<mode>): Adjust.
14092 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
14094 (rs6000_cpu_cpp_builtins): Adjust.
14095 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
14096 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
14097 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
14098 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
14099 -maltivec=be support.
14100 (rs6000_split_vec_extract_var): Adjust.
14101 (rs6000_split_v4si_init): Adjust.
14102 (swap_selector_for_mode): Delete.
14103 (altivec_expand_lvx_be, altivec_expand_stvx_be,
14104 altivec_expand_stvex_be): Delete.
14105 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
14106 -maltivec=be support.
14107 (rs6000_gimple_fold_builtin): Ditto.
14108 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
14110 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
14111 (TARGET_DIRECT_MOVE_64BIT): Adjust.
14112 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
14113 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
14114 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
14115 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
14116 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
14117 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
14118 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
14119 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
14120 anonymous split): Adjust.
14121 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
14122 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
14124 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
14126 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
14127 when --with-gxx-include-dir is also specified.
14128 * configure: Regenerate.
14130 2018-05-09 Jim Wilson <jimw@sifive.com>
14133 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
14134 * config/riscv/t-withmultilib: New.
14135 * config/riscv/withmultilib.h: New.
14136 * doc/install.texi: Document RISC-V --with-multilib-list support.
14138 2018-05-09 Richard Biener <rguenther@suse.de>
14140 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
14142 (vect_bb_vectorization_profitable_p): Adjust. Compute
14143 actual scalar cost using the cost vector and the add_stmt_cost
14146 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14148 PR rtl-optimization/85645
14149 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
14150 in the REG_CFA_REGISTER note for LR, don't leave it empty.
14152 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14154 PR rtl-optimization/85645
14155 * shrink-wrap.c (spread_components): Return a boolean saying if
14156 anything was changed.
14157 (try_shrink_wrapping_separate): Iterate spread_components until
14158 nothing changes anymore.
14160 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14162 PR rtl-optimization/85645
14163 * regrename.c (build_def_use): Also kill the chains that include the
14164 destination of a REG_CFA_REGISTER note.
14166 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14168 PR rtl-optimization/85645
14169 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
14170 insn that has a REG_CFA_REGISTER note.
14172 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
14174 * cfgexpand.c (expand_clobber): New function.
14175 (expand_gimple_stmt_1): Use it.
14176 * tree-vect-stmts.c (vect_clobber_variable): New function,
14178 (vectorizable_simd_clone_call): ...here.
14179 (vectorizable_store): Emit a clobber either side of an
14180 IFN_STORE_LANES sequence.
14181 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
14183 2018-05-09 Tom de Vries <tom@codesourcery.com>
14186 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
14187 (define_insn "trap_if_false"): Add exit after trap.
14189 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
14191 PR rtl-optimization/85638
14192 * bb-reorder.c: Include common/common-target.h.
14193 (create_forwarder_block): New function extracted from...
14194 (fix_up_crossing_landing_pad): ...here. Rename into...
14195 (dw2_fix_up_crossing_landing_pad): ...this.
14196 (sjlj_fix_up_crossing_landing_pad): New function.
14197 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
14198 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
14199 from both partitions and exit the loop after one iteration.
14201 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14204 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14206 (Basic PowerPC Built-in Functions): The new name of the
14207 subsection previously known as "PowerPC Built-in Functions".
14208 (Basic PowerPC Built-in Functions Available on all Configurations):
14210 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14212 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14214 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14216 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14219 2018-05-08 Jim Wilson <jimw@sifive.com>
14221 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
14222 (LD_EMUL_SUFFIX): New.
14223 (LINK_SPEC): Use it.
14225 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14227 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14229 (Basic PowerPC Built-in Functions): The new name of the
14230 subsection previously known as "PowerPC Built-in Functions".
14231 (Basic PowerPC Built-in Functions Available on all Configurations):
14233 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14235 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14237 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14239 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14242 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14245 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
14246 after cmpelim optimization.
14248 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
14250 * config.gcc: Support "goldmont".
14251 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
14252 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14253 PROCESSOR_GOLDMONT.
14254 * config/i386/i386.c (m_GOLDMONT): Define.
14255 (processor_target_table): Add "goldmont".
14256 (PTA_GOLDMONT): Define.
14257 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
14258 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
14259 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
14260 (fold_builtin_cpu): Add "goldmont".
14261 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
14262 (ix86_option_override_internal): Add "goldmont".
14263 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
14264 (processor_type): Add PROCESSOR_GOLDMONT.
14265 * config/i386/i386.md: Add CPU "glm".
14266 * config/i386/glm.md: New file.
14267 * config/i386/x86-tune.def: Add m_GOLDMONT.
14268 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
14270 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14273 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
14275 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
14276 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
14277 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
14280 * config/i386/i386.c (ix86_fold_builtin): Handle
14281 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
14284 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
14285 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
14287 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
14290 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
14291 (check_arch): Likewise.
14292 (check_fpu): Return the result rather than printing it.
14293 (end arch): Fix operator precedence.
14294 (end cpu): Likewise.
14295 (END): Print the result from check_fpu.
14297 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14298 Alan Hayward <alan.hayward@arm.com>
14299 David Sherwood <david.sherwood@arm.com>
14301 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
14302 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
14303 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
14304 (*fcmuo<mode>_and): New patterns.
14306 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14308 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
14309 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
14310 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
14311 (cmp_op, sve_imm_con): New code attributes.
14312 (SVE_COND_INT_CMP, imm_con): Delete.
14313 (cmp_op): Remove above unspecs from int attribute.
14314 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
14316 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
14317 comparison-specific unspecs.
14318 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14319 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
14320 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
14321 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
14322 (*vec_fcm<cmp_op><mode>): Rename to...
14323 (*fcm<cmp_op><mode>): ...this and adjust likewise.
14324 (*vec_fcmuo<mode>): Rename to...
14325 (*fcmuo<mode>): ...this and adjust likewise.
14326 (*pred_fcm<cmp_op><mode>): New pattern.
14327 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
14328 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
14330 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
14332 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
14333 (aarch64_emit_sve_predicated_cond): New function.
14334 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
14335 (aarch64_emit_unspec_cond_or): Replace with...
14336 (aarch64_emit_sve_or_conds): ...this new function. Use
14337 aarch64_emit_sve_ptrue_op for the individual comparisons and
14338 aarch64_emit_binop to OR them together.
14339 (aarch64_emit_inverted_unspec_cond): Replace with...
14340 (aarch64_emit_sve_inverted_cond): ...this new function. Use
14341 aarch64_emit_sve_ptrue_op for the comparison and
14342 aarch64_emit_unop to invert the result.
14343 (aarch64_expand_sve_vec_cmp_float): Update after the above
14344 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
14346 2018-05-07 Nathan Sidwell <nathan@acm.org>
14348 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
14349 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
14350 (Backwards Compatibility): Likewise.
14352 2018-05-07 Luis Machado <luis.machado@linaro.org>
14356 2018-05-07 Luis Machado <luis.machado@linaro.org>
14358 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14359 <prefetch_dynamic_strides>: New const bool field.
14360 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14361 prefetch_dynamic_strides.
14362 (exynosm1_prefetch_tune): Likewise.
14363 (thunderxt88_prefetch_tune): Likewise.
14364 (thunderx_prefetch_tune): Likewise.
14365 (thunderx2t99_prefetch_tune): Likewise.
14366 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14368 (aarch64_override_options_internal): Update to set
14369 PARAM_PREFETCH_DYNAMIC_STRIDES.
14370 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14371 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14372 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14373 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14374 prefetch-dynamic-strides setting.
14376 2018-05-07 Luis Machado <luis.machado@linaro.org>
14378 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14379 <minimum_stride>: New const int field.
14380 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14381 minimum_stride field.
14382 (exynosm1_prefetch_tune): Likewise.
14383 (thunderxt88_prefetch_tune): Likewise.
14384 (thunderx_prefetch_tune): Likewise.
14385 (thunderx2t99_prefetch_tune): Likewise.
14386 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14387 (aarch64_override_options_internal): Update to set
14388 PARAM_PREFETCH_MINIMUM_STRIDE.
14389 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14390 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14391 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14392 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14393 stride is constant and is below the minimum stride threshold.
14395 2018-05-07 Luis Machado <luis.machado@linaro.org>
14397 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
14400 2018-05-07 Luis Machado <luis.machado@linaro.org>
14402 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14403 <prefetch_dynamic_strides>: New const bool field.
14404 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14405 prefetch_dynamic_strides.
14406 (exynosm1_prefetch_tune): Likewise.
14407 (thunderxt88_prefetch_tune): Likewise.
14408 (thunderx_prefetch_tune): Likewise.
14409 (thunderx2t99_prefetch_tune): Likewise.
14410 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14412 (aarch64_override_options_internal): Update to set
14413 PARAM_PREFETCH_DYNAMIC_STRIDES.
14414 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14415 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14416 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14417 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14418 prefetch-dynamic-strides setting.
14420 2018-05-07 Luis Machado <luis.machado@linaro.org>
14422 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14423 <minimum_stride>: New const int field.
14424 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14425 minimum_stride field.
14426 (exynosm1_prefetch_tune): Likewise.
14427 (thunderxt88_prefetch_tune): Likewise.
14428 (thunderx_prefetch_tune): Likewise.
14429 (thunderx2t99_prefetch_tune): Likewise.
14430 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14431 (aarch64_override_options_internal): Update to set
14432 PARAM_PREFETCH_MINIMUM_STRIDE.
14433 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14434 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14435 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14436 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14437 stride is constant and is below the minimum stride threshold.
14439 2018-05-06 Jakub Jelinek <jakub@redhat.com>
14442 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14443 the type is addressable. Don't force op into register if it has
14446 2018-05-05 Roland McGrath <mcgrathr@google.com>
14449 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14450 any section for which we don't know a specific type it should have,
14451 regardless of name. Previously this was done only for the exact
14452 names ".init_array", ".fini_array", and ".preinit_array".
14453 (default_elf_asm_named_section): Add comment about
14454 relationship with default_section_type_flags and SECTION_NOTYPE.
14455 (get_section): Don't consider it a type conflict if one side has
14456 SECTION_NOTYPE and the other doesn't, as long as neither has the
14457 SECTION_BSS et al used in the default_section_type_flags logic.
14459 2018-05-05 Tom de Vries <tom@codesourcery.com>
14462 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14463 (workaround_barsyncs): New function.
14464 (nvptx_reorg): Use workaround_barsyncs.
14465 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14466 (define_expand "nvptx_membar_cta"): New define_expand.
14467 (define_insn "*nvptx_membar_cta"): New insn.
14469 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
14471 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14472 To improve optimization opportunities.
14473 * builtin-types.def: The new needed builtin types for the above.
14475 2018-05-04 Richard Biener <rguenther@suse.de>
14477 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14478 * gimple-ssa-store-merging.c
14479 (imm_store_chain_info::output_merged_store): Remove redundant create,
14480 release split_store vector contents on failure.
14481 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14482 scalar stmt vector on cache hit.
14484 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14486 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14488 * config.gcc (powerpc-xilinx-eabi*): Remove.
14489 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14491 (fusion_addis_mem_combo_load): Ditto.
14492 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14494 (rs6000_cpu_cpp_builtins): Ditto.
14495 * config/rs6000/rs6000-linux.c
14496 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14497 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14498 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14500 (rs6000_setup_reg_addr_masks): Ditto.
14501 (rs6000_init_hard_regno_mode_ok): Ditto.
14502 (rs6000_option_override_internal): Ditto.
14503 (legitimate_lo_sum_address_p): Ditto.
14504 (rs6000_legitimize_address): Ditto.
14505 (rs6000_legitimize_reload_address): Ditto.
14506 (rs6000_legitimate_address_p): Ditto.
14507 (abi_v4_pass_in_fpr): Ditto.
14508 (setup_incoming_varargs): Ditto.
14509 (rs6000_gimplify_va_arg): Ditto.
14510 (rs6000_split_multireg_move): Ditto.
14511 (rs6000_savres_strategy): Ditto.
14512 (rs6000_emit_prologue_components): Ditto.
14513 (rs6000_emit_epilogue_components): Ditto.
14514 (rs6000_emit_prologue): Ditto.
14515 (rs6000_emit_epilogue): Ditto.
14516 (rs6000_elf_file_end): Ditto.
14517 (rs6000_function_value): Ditto.
14518 (rs6000_libcall_value): Ditto.
14519 * config/rs6000/rs6000.h: Ditto.
14520 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14521 (TARGET_MINMAX): ... this. New.
14522 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14523 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14524 (*movsi_internal1_single): Delete.
14525 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14526 mfpu=, mxilinx-fpu): Delete.
14527 * config/rs6000/singlefp.h: Delete.
14528 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14529 * config/rs6000/t-rs6000: Ditto.
14530 * config/rs6000/t-xilinx: Delete.
14531 * config/rs6000/titan.md: Adjust for fp_type removal.
14532 * config/rs6000/vsx.md: Remove Xilinx FP support.
14533 (VStype_simple): Delete.
14534 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14535 * config/rs6000/xfpu.h: Delete.
14536 * config/rs6000/xfpu.md: Delete.
14537 * config/rs6000/xilinx.h: Delete.
14538 * config/rs6000/xilinx.opt: Delete.
14539 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14540 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14542 2018-05-04 Tom de Vries <tom@codesourcery.com>
14545 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14548 2018-05-04 Richard Biener <rguenther@suse.de>
14550 PR middle-end/85627
14551 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14552 (expand_complex_div_wide): Likewise.
14553 (expand_complex_operations_1): Likewise.
14554 (expand_complex_libcall): Preserve EH info of the original stmt.
14555 (tree_lower_complex): Handle removed blocks.
14556 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14557 on complex multiplication and division libcall builtins.
14559 2018-05-04 Richard Biener <rguenther@suse.de>
14561 PR middle-end/85574
14562 * fold-const.c (negate_expr_p): Restrict negation of operand
14563 zero of a division to when we know that can happen without
14565 (fold_negate_expr_1): Likewise.
14567 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14570 * real.h (real_nextafter): Declare.
14571 * real.c (real_nextafter): New function.
14572 * fold-const-call.c (fold_const_nextafter): New function.
14573 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14574 CASE_CFN_NEXTTOWARD.
14575 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14576 even when arg1_mode is different from arg0_mode.
14578 2018-05-03 Nathan Sidwell <nathan@acm.org>
14580 * doc/extend.texi (Deprecated Features): Remove
14581 -ffriend-injection.
14582 (Backwards Compatibility): Likewise.
14583 * doc/invoke.texi (C++ Language Options): Likewise.
14584 (C++ Dialect Options): Likewise.
14586 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14589 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14590 _mm512_mask_mullox_epi64): New intrinsics.
14592 2018-05-03 Tom de Vries <tom@codesourcery.com>
14595 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14596 dump files): Add offload-tree.
14598 2018-05-03 Richard Biener <rguenther@suse.de>
14600 PR tree-optimization/85615
14601 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14602 to loops not nested in BBs loop father to avoid creating multi-entry
14605 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14607 PR tree-optimization/70291
14608 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14609 arguments. Change return type to tree. Emit libcall as a new
14610 statement rather than replacing existing one when inplace_p is true.
14611 (expand_complex_multiplication_components): New function.
14612 (expand_complex_multiplication): Expand floating-point complex
14613 multiplication using the above.
14614 (expand_complex_division): Rename inner_type parameter to type.
14615 Update expand_complex_libcall call-site.
14616 (expand_complex_operations_1): Update expand_complex_multiplication
14617 and expand_complex_division call-sites.
14619 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14622 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14623 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14624 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14625 the highest significant bit of the shift count mask is clear. In
14626 check whether and[sq]i3 is needed verify that all significant bits
14627 of the shift count other than the highest are set.
14629 2018-05-02 Tom de Vries <tom@codesourcery.com>
14632 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14633 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14634 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14635 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14636 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14637 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14638 __builtin_goacc_parlevel_size.
14640 2018-05-02 Richard Biener <rguenther@suse.de>
14642 PR tree-optimization/85597
14643 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14644 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14647 2018-05-02 Tom de Vries <tom@codesourcery.com>
14650 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14651 dump files): Add ltrans-tree.
14653 2018-05-02 Tom de Vries <tom@codesourcery.com>
14656 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14657 dump files): Add wpa-ipa.
14659 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14661 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14662 powerpc*-*-linux*paired* target.
14663 * config/rs6000/750cl.h: Delete.
14664 * config/rs6000/paired.h: Delete.
14665 * config/rs6000/paired.md: Delete.
14666 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14668 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14669 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14670 comment. Remove paired float support.
14671 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14672 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14674 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14675 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14677 * config/rs6000/rs6000.c: Remove paired float support.
14678 (paired_expand_vector_init, paired_expand_vector_move,
14679 paired_emit_vector_compare, paired_emit_vector_cond_expr,
14680 (paired_expand_lv_builtin, paired_expand_stv_builtin,
14681 paired_expand_builtin, paired_expand_predicate_builtin,
14682 paired_init_builtins): Delete.
14683 * config/rs6000/rs6000.h: Remove paired float support.
14684 * config/rs6000/rs6000.md: Remove paired float support.
14685 (move_from_CR_ov_bit): Delete.
14686 * config/rs6000/rs6000.opt (mpaired): Delete.
14687 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14688 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14690 2018-05-02 Richard Biener <rguenther@suse.de>
14692 PR middle-end/85567
14693 * gimplify.c (gimplify_save_expr): When in SSA form allow
14694 SAVE_EXPRs to compute to SSA vars.
14696 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14699 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14700 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14701 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14702 clobber operands[2], instead use a new pseudo. Formatting fixes.
14704 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
14706 PR tree-optimization/85586
14707 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14708 exit early for statements in the same group if the accesses are
14711 2018-05-02 Tom de Vries <tom@codesourcery.com>
14714 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14717 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
14719 PR tree-optimization/85143
14720 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14722 2018-05-01 Tom de Vries <tom@codesourcery.com>
14725 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14726 not found" error message.
14728 2018-05-01 Tom de Vries <tom@codesourcery.com>
14731 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14732 * vec.c (test_ordered_remove_if): New function.
14733 (vec_c_tests): Call test_ordered_remove_if.
14734 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14735 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14736 * tree-vect-patterns.c (vect_pattern_recog_1): Use
14737 VEC_ORDERED_REMOVE_IF.
14739 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14741 PR tree-optimization/82665
14742 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14743 pointer subtraction where arguments come from a memchr call.
14745 2018-05-01 Jakub Jelinek <jakub@redhat.com>
14747 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14748 --push-state --as-needed and --pop-state instead of --as-needed and
14749 --no-as-needed if ld supports it.
14750 * configure: Regenerated.
14753 * doc/install.texi2html: Replace _002d with - and _002a with * in
14754 generated html files using sed.
14756 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14759 * gcc-rich-location.c (blank_line_before_p): New function.
14760 (use_new_line): New function.
14761 (gcc_rich_location::add_fixit_insert_formatted): New function.
14762 * gcc-rich-location.h
14763 (gcc_rich_location::add_fixit_insert_formatted): New function.
14765 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14767 * selftest.c (assert_streq): Rename "expected" and "actual" to
14768 "val1" and "val2". Extend NULL-handling to cover both inputs
14769 symmetrically, while still requiring both to be non-NULL for a pass.
14770 * selftest.h (assert_streq): Rename "expected" and "actual" to
14772 (ASSERT_EQ): Likewise.
14773 (ASSERT_EQ_AT): Likewise.
14774 (ASSERT_KNOWN_EQ): Likewise.
14775 (ASSERT_KNOWN_EQ_AT): Likewise.
14776 (ASSERT_NE): Likewise.
14777 (ASSERT_MAYBE_NE): Likewise.
14778 (ASSERT_MAYBE_NE_AT): Likewise.
14779 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
14780 the assertion to pass.
14781 (ASSERT_STREQ_AT): Likewise.
14783 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14785 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14786 interaction with -pie.
14788 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14790 * selftest.h: Fix alphabetization of per-source-file selftest
14793 2018-04-30 Jason Merrill <jason@redhat.com>
14795 PR c++/61982 - dead stores to destroyed objects.
14796 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14799 2018-04-30 Jason Merrill <jason@redhat.com>
14801 * tree.c (build_clobber): New.
14802 * tree.h: Declare it.
14803 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14805 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14807 * diagnostic-show-locus.c (layout::layout): Update for
14808 location_get_source_line returning a char_span.
14809 (struct char_span): Move to input.h.
14810 (struct correction): Update for fields in char_span becoming
14812 (struct source_line): Update for location_get_source_line
14813 returning a char_span.
14814 (layout::print_line): Likewise.
14815 * edit-context.c (edited_file::print_content): Likewise.
14816 (edited_file::print_diff_hunk): Likewise.
14817 (edited_file::print_run_of_changed_lines): Likewise.
14818 (edited_file::get_num_lines): Likewise.
14819 (edited_line::edited_line): Likewise.
14820 * final.c (asm_show_source): Likewise.
14821 * input.c (location_get_source_line): Convert return type
14822 from const char * to char_span, losing the final "line_len"
14824 (dump_location_info): Update for the above.
14825 (get_substring_ranges_for_loc): Likewise. Use a char_span
14826 when handling the literal within the line.
14827 (test_reading_source_line): Update for location_get_source_line
14828 returning a char_span.
14829 * input.h (class char_span): Move here from
14830 diagnostic-show-locus.c, converting from a struct to a class.
14831 Make data members private.
14832 (char_span::operator bool): New.
14833 (char_span::length): New.
14834 (char_span::get_buffer): New.
14835 (char_span::operator[]): New.
14836 (char_span::subspan): Make const.
14837 (char_span::xstrdup): New.
14838 (location_get_source_line): Convert return type from const char *
14839 to char_span, losing the final "line_size" param.
14841 2018-04-30 Jan Hubicka <jh@suse.cz>
14843 * lto-wrapper.c (ltrans_priorities): New static var.
14844 (cmp_priority): New.
14845 (run_gcc): Read priorities and if doing parallel build order
14846 the Makefile by them.
14848 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14850 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14852 2018-04-30 Richard Biener <rguenther@suse.de>
14854 * tree-cfg.c (verify_address): Remove base argument, add
14855 flag whether to check TREE_ADDRESSABLE and do that.
14856 (verify_expr): Remove.
14857 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14858 (verify_gimple_assign_single): Likewise.
14859 (verify_gimple_switch): Likewise.
14860 (verify_expr_location_1): Dereference tp once. Add (disabled)
14861 piece from verify_expr.
14862 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14864 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14866 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14868 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14870 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14871 (small_data_pattern): Likewise.
14872 (arc_rewrite_small_data): Likewise.
14873 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14874 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14875 (get_symbol_alignment): New function.
14876 (legitimate_small_data_address_p): Likewise.
14877 (legitimate_scaled_address): Update, call
14878 legitimate_small_data_address_p.
14879 (output_sdata): New static variable.
14880 (arc_print_operand): Update how we handle small data operands.
14881 (arc_print_operand_address): Likewise.
14882 (arc_legitimate_address_p): Update, use
14883 legitimate_small_data_address_p.
14884 (arc_rewrite_small_data_p): Remove.
14885 (arc_rewrite_small_data_1): Likewise.
14886 (arc_rewrite_small_data): Likewise.
14887 (small_data_pattern): Likewise.
14888 (compact_sda_memory_operand): Update to use
14889 legitimate_small_data_address_p and get_symbol_alignment.
14890 (prepare_move_operands): Don't rewite sdata pattern.
14891 (prepare_extend_operands): Remove.
14892 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14894 (zero_extendqisi2): Likewise.
14895 (zero_extendhisi2): Likewise.
14896 (extendqihi2): Likewise.
14897 (extendqisi2): Likewise.
14898 (extendhisi2): Likewise.
14899 (addsi3): Likewise.
14900 (subsi3): Likewise.
14901 (andsi3): Likewise.
14902 * config/arc/constraints.md (Usd): Change it to memory constraint.
14904 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14906 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14907 as source of std instructions.
14908 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14909 allow 6-bit constants as source for store instructions.
14910 (movdi_insn): Update instruction pattern to allow 6-bit constants
14911 as source for store instructions.
14913 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14915 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14917 2018-04-30 Nathan Sidwell <nathan@acm.org>
14918 Sandra Loosemore <sandra@codesourcery.com>
14920 * dumpfile.c (dump_open): Allow '-' for stdout.
14921 * doc/invoke.texi (Developer Options): Document dump filename
14922 determination early. Document stdin/stdout selection.
14924 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14926 Microblaze Target: PIC data text relative
14928 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14929 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14931 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14932 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14933 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14934 New addressing mode for data-text relative position indepenedent code.
14935 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14936 'ADDRESS_SYMBOLIC_TXT_REL'.
14937 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14938 (microblaze_legitimate_pic_operand): Exclude function calls from
14939 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14940 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14942 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14943 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14944 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14945 for 'address + offset'.
14946 (microblaze_expand_prologue): Add new function prologue call for
14948 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14949 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14950 table in case of TARGET_PIC_DATA_TEXT_REL.
14951 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14952 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14953 Add new macros 'UNSPEC_TEXT',
14954 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14955 + exclude function calls from 'UNSPEC_PLT' in case of data text
14957 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14958 new target hook for generating address diff vector tables in case of
14960 * doc/tm.texi : Regenerate.
14961 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14962 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14963 of addr diff vector generation.
14964 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14965 target hook definition.
14966 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14967 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14968 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14969 option for data text relative.
14971 2018-04-30 Richard Biener <rguenther@suse.de>
14973 * tree-chrec.h (evolution_function_is_constant_p): Remove
14975 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14977 2018-04-30 Richard Biener <rguenther@suse.de>
14980 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14982 2018-04-30 Richard Biener <rguenther@suse.de>
14984 PR tree-optimization/28364
14985 PR tree-optimization/85275
14986 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14987 copying first exit test.
14989 2018-04-28 Mark Wielaard <mark@klomp.org>
14991 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14992 dwarf_version >= 5.
14993 (dwarf_AT): Handle DW_AT_addr_base.
14994 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14996 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14999 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
15000 (*ashl<dwi>3_doubleword_mask_1): Ditto.
15001 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
15002 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
15004 2018-04-28 Richard Biener <rguenther@suse.de>
15006 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
15007 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
15008 to reflect use. Only add interesting stmts.
15010 2018-04-27 Martin Jambor <mjambor@suse.cz>
15013 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
15014 the jump function allows for passing through aggregate values.
15016 2018-04-27 David Malcolm <dmalcolm@redhat.com>
15018 * input.h (in_system_header_at): Convert from macro to inline
15020 (from_macro_expansion_at): Likewise.
15021 (from_macro_definition_at): Likewise.
15023 2018-04-27 Jeff Law <law@redhat.com>
15025 * config.gcc: Mark tile* targets as deprecated/obsolete.
15027 2018-04-27 Richard Biener <rguenther@suse.de>
15029 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
15032 2018-04-27 Richard Biener <rguenther@suse.de>
15034 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
15036 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
15038 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
15039 with Yd constraint. Set "preferred_for_speed" attribute from
15040 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
15041 with Yd constraint.
15042 (*movdi_internal): Ditto.
15043 (movti_interunit splitters): Remove
15044 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
15045 (movdi_interunit splitters): Ditto.
15046 * config/i386/constraints.md (Ye): Remove.
15047 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
15049 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15052 * config/aarch64/constraints.md (Usg): Limit to 31.
15053 (Usj): Limit to 63.
15055 2018-04-27 Jakub Jelinek <jakub@redhat.com>
15057 PR tree-optimization/85529
15058 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
15059 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
15060 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
15061 zero extension or masking of the MSB bit.
15062 (optimize_range_tests): Add FIRST_BB argument, pass it through
15063 to optimize_range_tests_var_bound.
15064 (maybe_optimize_range_tests, reassociate_bb): Adjust
15065 optimize_range_tests callers.
15067 2018-04-26 Richard Biener <rguenther@suse.de>
15068 Jakub Jelinek <jakub@redhat.com>
15070 * cgraph.h (symbol_table): Just declare debug method here.
15071 * symtab.c (symbol_table::debug): Define.
15073 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
15075 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
15077 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
15079 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
15080 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
15081 (*movdi_internal): Substitute Yi and Yj constraint with x
15082 and Ym and Yn constraint with y constraint. Update "isa"
15083 attribute and set "preferred_for_speed" attribute from
15084 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
15085 (*movsi_internal): Ditto.
15086 (*movdf_internal): Ditto.
15087 (*movsf_internal): Ditto.
15088 (*zero_extendsidi2): Ditto.
15089 * config/i386/sse.md (vec_set<mode>_0): Ditto.
15090 (sse2_loadld): Ditto.
15091 (*vec_extract<ssevecmodelower>_0): Ditto.
15092 (*vec_extractv4si_0_zext_sse4): Ditto.
15093 (vec_concatv2di): Ditto.
15094 (*vec_dup<mode>): Ditto.
15095 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
15096 * config/i386/constraints.md (Yi): Remove.
15101 2018-04-26 Nathan Sidwell <nathan@acm.org>
15103 * dumpfile.c (dump_open): New.
15104 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
15105 (dump_finish): Detect stdio/stderr by value not name.
15107 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
15109 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
15111 2018-04-26 Tom de Vries <tom@codesourcery.com>
15114 * config/nvptx/nvptx.c (verify_neutering_jumps)
15115 (verify_neutering_labels): New function
15116 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
15118 2018-04-26 Tom de Vries <tom@codesourcery.com>
15121 * config/nvptx/nvptx.c (needs_neutering_p): New function.
15122 (nvptx_single): Use needs_neutering_p to skip over insns that do not
15125 2018-04-26 Richard Biener <rguenther@suse.de>
15126 Tom de Vries <tom@codesourcery.com>
15129 * lto-streamer-out.c (output_function): Fixup loops if required to match
15130 discovery done in the reader.
15132 2018-04-26 Richard Biener <rguenther@suse.de>
15134 PR tree-optimization/85116
15135 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
15136 have a loop exit from the single latch predecessor. Remove
15137 case of header with just condition.
15138 (ch_base::copy_headers): Exclude infinite loops from any
15140 (pass_ch::execute): Record exits.
15142 2018-04-26 Richard Biener <rguenther@suse.de>
15144 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
15145 prologue cost vector and pass it to vect_get_load_cost.
15146 (vect_get_peeling_costs_all_drs): Likewise.
15147 (vect_peeling_hash_get_lowest_cost): Likewise.
15148 (vect_enhance_data_refs_alignment): Likewise.
15150 2018-04-26 Richard Biener <rguenther@suse.de>
15152 PR middle-end/85450
15153 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
15154 checking of integer<->pointer conversions.
15155 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
15156 sign-/zero-extending pointer types.
15157 (expand_omp_for_static_chunk): Likewise.
15159 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
15160 Jean Lee <xiaoyur347@gmail.com>
15162 * config/mips/mips.c (mips_asan_shadow_offset): New function.
15163 (TARGET_ASAN_SHADOW_OFFSET): Define.
15164 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
15165 true for -fsanitize=address.
15167 2018-04-25 Mark Wielaard <mark@klomp.org>
15169 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
15172 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15174 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
15175 than "alu", remove explicit "memory" and "imm_disp" attributes.
15176 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
15178 PR middle-end/85414
15179 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
15180 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
15181 gen_lowpart_no_emit.
15183 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
15186 * config/i386/i386.c (ix86_expand_builtin): Change memory
15187 operand to XI, extend p0 to Pmode.
15188 * config/i386/i386.md: Change unspec volatile and operand
15189 1 mode to XI, change operand 0 mode to P.
15191 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15193 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
15194 GET_MODE_MASK before any checking.
15195 (nds32_can_use_bset_p): Likewise.
15196 (nds32_can_use_btgl_p): Likewise.
15198 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15200 * config/nds32/nds32-doubleword.md: New define_split pattern for
15201 illegal register number.
15203 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15205 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
15207 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15209 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
15211 2018-04-25 Richard Biener <rguenther@suse.de>
15213 * lto-streamer.h (LTO_major_version): Bump to 8.
15215 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15217 * BASE-VER: Set to 9.0.0.
15219 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
15221 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
15222 in __abskf2 and __powikf2.
15224 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15227 * config/aarch64/constraints.md (Usg, Usj): New constraints.
15228 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
15229 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
15230 Use the above on operand 2. Reindent.
15231 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15233 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15236 * common/config/i386/i386-common.c (ix86_handle_option): Don't
15238 * config/i386/i386.opt (mcet): Removed.
15239 * doc/install.texi: Remove -mcet documentation.
15240 * doc/invoke.texi: Likewise.
15242 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15245 * doc/install.texi: Remove -mcet from bootstrap-cet.
15247 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15250 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
15251 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
15255 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
15256 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
15257 containing a CONST_VECTOR.
15259 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
15261 * doc/install.texi: Update newlib dependency for nvptx.
15263 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15266 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
15267 instead of INTVAL when shifting x left.
15269 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
15271 PR tree-optimization/85478
15272 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
15273 vect_grouped_store_supported for single element vectors.
15275 2018-04-24 Richard Biener <rguenther@suse.de>
15278 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
15279 load cost increase to the case of non-constant step.
15281 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15284 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
15285 destination if any_malformed_asm.
15287 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
15289 PR middle-end/85496
15290 * expr.c (store_field): In the bitfield case, if the value comes from
15291 a function call and is returned in registers by means of a PARALLEL,
15292 do not change the mode of the temporary unless BLKmode and VOIDmode.
15294 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
15296 PR rtl-optimization/85423
15297 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
15298 dependencies to debug insns when the previous insn is non-debug.
15300 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15302 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
15303 enums into a single definition.
15304 (fls): Fix predicates and printing.
15307 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15309 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
15310 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
15311 and short u6 immediate.
15312 (check_if_valid_sleep_operand): Remove.
15313 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
15315 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15317 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
15318 flag_always_save_lp condition.
15319 * config/nds32/nds32.opt (malways-save-lp): New option.
15321 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15323 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
15324 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
15325 * config/nds32/nds32.h
15326 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
15327 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
15329 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15331 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
15332 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
15334 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15335 Chung-Ju Wu <jasonwucj@gmail.com>
15337 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
15339 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
15340 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
15342 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15344 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
15346 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15348 * config/nds32/nds32-protos.h (nds32_data_alignment,
15349 nds32_local_alignment): Declare.
15350 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
15351 nds32_local_alignment): New functions.
15352 (TARGET_CONSTANT_ALIGNMENT): Define.
15353 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
15355 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15357 * config/nds32/nds32.c
15358 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
15359 (TARGET_MODES_TIEABLE_P): Likewise.
15361 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15363 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
15364 level Ofast and Og.
15366 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
15367 Chung-Ju Wu <jasonwucj@gmail.com>
15369 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15370 for unaligned access.
15371 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15373 * config/nds32/nds32-intrinsic.md: Likewise.
15374 * config/nds32/nds32_intrinsic.h: Likewise.
15375 * config/nds32/nds32.h (nds32_builtins): Likewise.
15376 * config/nds32/nds32.opt (munaligned-access): New option.
15377 * config/nds32/nds32.c (nds32_asm_file_start): Display
15378 flag_unaligned_access status.
15380 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
15382 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
15383 -mno-relax is present.
15384 * config/riscv/linux.h (LINK_SPEC): Ditto.
15386 2018-04-20 Martin Sebor <msebor@redhat.com>
15389 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
15391 (gimple_fold_builtin_stxcpy_chk): Same.
15392 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
15394 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
15397 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
15398 __powikf2 when long double is IEEE 128-bit.
15400 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
15402 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
15403 step to make sure stack always aligned.
15405 2018-04-20 Carl Love <cel@us.ibm.com>
15408 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
15409 size check for arg0.
15411 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
15412 Tom de Vries <tom@codesourcery.com>
15415 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
15416 Emit insns for calls too.
15417 (nvptx_find_par): Always look for worker-level predecessor insn.
15418 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
15420 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
15421 (nvptx_process_pars): Propagate frames for calls.
15423 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
15426 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
15428 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
15429 (ix86_handle_option): Don't handle OPT_mibt.
15430 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
15432 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
15434 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
15435 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
15436 (ix86_target_macros): Define __CET__ with flag_cf_protection
15437 for -fcf-protection.
15438 * config/i386/i386.c (isa2_opts): Remove -mibt.
15439 * config/i386/i386.h (TARGET_IBT): Removed.
15440 (TARGET_IBT_P): Likewise.
15441 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
15442 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15443 * config/i386/i386.opt (mcet): Update help message.
15444 (mshstk): Likewise.
15446 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
15447 -mcet as an alias for -mshstk.
15449 2018-04-20 Richard Biener <rguenther@suse.de>
15451 PR middle-end/85475
15452 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15453 complexity by forcing a single use of the multiply operand.
15455 2018-04-20 Martin Jambor <mjambor@suse.cz>
15458 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15459 recursion dependency to only apply to non-clones.
15461 2018-04-20 Martin Jambor <mjambor@suse.cz>
15464 * ipa-cp.c (create_specialized_node): Check that clones of
15465 self-recursive edges exist during IPA-CP.
15467 2018-04-19 Toon Moene <toon@moene.org>
15469 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15472 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15474 PR tree-optimization/85467
15475 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15476 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
15477 VECTOR_CST element to type.
15479 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15482 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15483 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15484 (builtin_longjmp): Likewise.
15485 (save_stack_nonlocal): New pattern.
15486 (restore_stack_nonlocal): Likewise.
15488 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15491 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15492 Replace ASM_OUTPUT_LABEL with fprintf.
15494 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15497 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15498 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15499 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15500 define __IBT__ and __SHSTK__ for -fcf-protection.
15501 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15503 (ix86_trampoline_init): Likewise.
15504 (x86_output_mi_thunk): Likewise.
15505 (ix86_notrack_prefixed_insn_p): Likewise.
15506 (ix86_option_override_internal): Don't disallow -fcf-protection.
15507 * config/i386/i386.md (rdssp<mode>): Also enable for
15509 (incssp<mode>): Likewise.
15510 (nop_endbr): Likewise.
15511 * config/i386/i386.opt (mcet): Change help message to built-in
15514 (mshstk): Likewise.
15515 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15516 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15517 enable CET built-in functions.
15519 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15521 * common/config/i386/i386-common.c
15522 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15523 OPTION_MASK_ISA_MOVDIRI_UNSET,
15524 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15525 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15526 * config.gcc (movdirintrin.h): New header.
15527 * config/i386/cpuid.h (bit_MOVDIRI,
15528 bit_MOVDIR64B): New bits.
15529 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15531 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15532 (VOID, PVOID, PCVOID)): New function types.
15533 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15534 __builtin_ia32_directstoreu_u64,
15535 __builtin_ia32_movdir64b): New builtins.
15536 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15537 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15539 (ix86_valid_target_attribute_inner_p): Ditto.
15540 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15541 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15542 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15543 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15544 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15545 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15546 (movdiri<mode>, movdir64b_<mode>): New.
15547 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15548 * config/i386/immintrin.h: Include movdirintrin.h.
15549 * config/i386/movdirintrin.h: New file.
15550 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15552 2018-04-19 Richard Biener <rguenther@suse.de>
15554 PR middle-end/85455
15555 * cfg.c (clear_bb_flags): When loop state says we have
15556 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15558 2018-04-19 Richard Biener <rguenther@suse.de>
15560 PR tree-optimization/84737
15561 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15562 copying restrict info.
15563 (vect_setup_realignment): Use it.
15564 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15565 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15566 the first DR to all generated stores.
15567 (vectorizable_load): Likewise for loads.
15569 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15571 PR tree-optimization/85446
15572 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15573 the integral and pointer types to have the same precision.
15575 * doc/install.texi: Document --disable-cet being the default and
15578 2018-04-18 Martin Liska <mliska@suse.cz>
15580 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15583 2018-04-18 Martin Liska <mliska@suse.cz>
15586 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15589 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15590 arguments if they are comparable.
15592 2018-04-18 Martin Liska <mliska@suse.cz>
15595 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15598 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15601 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15604 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15605 ENDBR after calling __morestack.
15607 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15610 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15611 by using gcc_base_ver to generate a gcc_driver_version, and use
15612 it when generating GCC_DRIVER_NAME.
15613 * configure: Regenerate.
15615 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15618 * config.gcc: Obsolete powerpc*-*-*spe*.
15620 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15623 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15624 (stabstr_D): Change type of unum from unsigned int to
15625 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15628 2018-04-17 Jim Wilson <jimw@sifive.com>
15631 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15632 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15633 Set arg_pointer_offset after using pretend_args_size.
15635 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15637 PR rtl-optimization/85431
15638 * dse.c (record_store): Ignore zero width stores.
15641 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15642 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15643 __builtin_stack_restore rather than after it.
15644 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15645 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15646 argument instead of virtual_dynamic_stack_rtx.
15648 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15650 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15652 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15653 Add note to error message to explain internal mapping of overloaded
15654 built-in function name to non-overloaded built-in function name.
15655 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15658 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15661 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15662 where the inputs overlap with the output.
15664 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15666 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15667 (=v, v) alternative and explicit "memory" attribute.
15668 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15669 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15671 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15672 "sselog1" type instead of "sselog".
15673 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15674 "sselog". Remove explicit "memory" attribute.
15675 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15676 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15678 (vec_extract_hi_v32hi): Merge all alternatives into one, use
15679 "sselog1" type instead of "sselog". Remove explicit "memory"
15681 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15682 use "sselog1" type instead of "sselog". Remove explicit "memory"
15684 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15685 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15687 (vec_extract_hi_v64qi): Merge all alternatives into one, use
15688 "sselog1" type instead of "sselog". Remove explicit "memory"
15690 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15691 use "sselog1" type instead of "sselog". Remove explicit "memory"
15695 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15697 PR middle-end/85414
15698 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15701 2018-04-17 Martin Jambor <mjambor@suse.cz>
15704 * ipa-cp.c (create_specialized_node): Call
15705 expand_all_artificial_thunks if necessary.
15707 2018-04-17 Martin Liska <mliska@suse.cz>
15710 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15711 in message, remote space in between '_G' and '('.
15713 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15716 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15717 avx512f_vmcmp<mode>3<round_saeonly_name>,
15718 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15719 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15720 avx512f_rndscale<mode><round_saeonly_name>,
15721 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15722 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15723 Use %<iptr>2 instead of %2 for -masm=intel.
15724 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15725 avx512f_vcvttss2usi<round_saeonly_name>,
15726 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15728 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15729 avx512f_vcvttsd2usi<round_saeonly_name>,
15730 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15731 Use %q1 instead of %1 for -masm=intel.
15732 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15733 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15734 of %3 for -masm=intel.
15735 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15737 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15739 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15741 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15742 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15744 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15746 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15747 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15748 %g1 and one with %0 and %1.
15749 (avx512er_vmrcp28<mode><round_saeonly_name>,
15750 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15751 %1 for -masm=intel.
15752 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15753 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15754 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15755 of %0 and %{%4%} for -masm=intel.
15756 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15757 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15758 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15759 order of %0 and %{%5%}%{z%} for -masm=intel.
15761 2018-04-17 Jan Hubicka <jh@suse.cz>
15764 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15766 2018-04-17 Martin Liska <mliska@suse.cz>
15769 * multiple_target.c (create_dispatcher_calls): Set apostrophes
15770 for target_clone error message. Make default implementation
15771 clone to be a local declaration.
15772 (separate_attrs): Add new argument and check for an empty
15774 (expand_target_clones): Handle it.
15775 (ipa_target_clone): Make redirection just for target_clones
15778 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
15779 Tom de Vries <tom@codesourcery.com>
15781 PR middle-end/84955
15782 * omp-expand.c (expand_oacc_for): Add dummy false branch for
15783 tiled basic blocks without omp continue statements.
15785 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
15788 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15789 vec_extract expression as having side effects to make sure it gets
15792 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15795 * config/i386/i386.c (get_builtin_code_for_version): Check
15798 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15801 * config.gcc: Support "skylake".
15802 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15804 * config/i386/i386.c (m_SKYLAKE): Define.
15805 (processor_target_table): Add "skylake".
15806 (ix86_option_override_internal): Add "skylake".
15807 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15808 PROCESSOR_CANNONLAKE.
15809 (get_builtin_code_for_version): Fix priority for
15810 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15811 PROCESSOR_SKYLAKE-AVX512.
15812 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15813 (processor_type): Add PROCESSOR_SKYLAKE.
15815 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15816 Jason Merrill <jason@redhat.com>
15819 * convert.c (convert_to_integer_1): Use direct recursion for
15820 enumeral types and types with a precision less than the number
15821 of bits in their mode.
15823 2018-04-16 Julia Koval <julia.koval@intel.com>
15826 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15827 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15829 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15832 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15833 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15834 and -mno-direct-move.
15836 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15839 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15840 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15841 Ensure negative shifts result in {0}.
15843 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15845 PR rtl-optimization/79916
15846 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15847 regs (if any) to define how to gnerate SD moves when LRA is in
15850 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15852 PR rtl-optimization/85393
15853 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15854 * except.c (expand_dw2_landing_pad_for_region): Make static.
15855 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15856 a label and unconditional jump to old_bb, rather than
15857 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15860 PR rtl-optimization/85376
15861 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15862 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15863 instead of a specific value.
15865 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15866 Bin Cheng <bin.cheng@arm.com>
15868 PR tree-optimization/82965
15869 PR tree-optimization/83991
15870 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15871 by_profile_only parameter.
15872 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15873 information if the loop was predicted to iterate too many times.
15874 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15876 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15879 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15882 2018-04-13 Martin Liska <mliska@suse.cz>
15883 Jakub Jelinek <jakub@redhat.com>
15885 PR middle-end/81657
15886 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15887 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15888 * builtins.c (expand_builtin_memory_copy_args): Use
15889 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15890 handle dest_addr == pc_rtx.
15892 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15895 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15896 asked to not generate direct moves.
15897 (fix_trunc<mode>si2_stfiwx): Similar.
15898 (fix_trunc<mode>si2_internal): Similar.
15900 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15903 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15904 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15905 lookup if dest in some wider mode is known to be const0_rtx and
15906 if so, record permanent equivalence for it to be ZERO_EXTEND of
15907 the narrower mode destination.
15909 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15911 * lto-streamer-out.c (output_function): Revert 259346.
15912 * omp-expand.c (expand_oacc_for): Likewise.
15914 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15916 PR rtl-optimization/85354
15917 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15918 * sel-sched.c (sel_global_init): ... here.
15920 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15923 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15924 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15925 mode for PE-COFF targets.
15926 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15927 (i386_pe_asm_lto_end): Likewise.
15928 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15929 (TARGET_ASM_LTO_END): Likewise.
15930 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15931 (i386_pe_asm_lto_start): New function.
15932 (i386_pe_asm_lto_end): Likewise.
15934 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15935 Richard Biener <rguenther@suse.de>
15937 PR middle-end/84955
15938 * lto-streamer-out.c (output_function): Fix CFG loop state before
15940 * omp-expand.c (expand_oacc_for): Handle calls to internal
15941 functions like regular functions.
15943 2018-04-12 Richard Biener <rguenther@suse.de>
15946 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15947 for the early LTO debug to properly generate references to it
15948 during DIE emission. Do not re-use that for the skeleton for
15950 (dwarf2out_early_finish): Likewise.
15952 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15955 * config/i386/sse.md
15956 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15957 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15958 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15959 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15960 and output is a reg, avoid creating invalid lowpart subreg, but
15961 instead split into a 512-bit move. Don't split if not AVX512VL,
15962 input is xmm16+ reg and output is a mem.
15963 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15964 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15965 xmm16+ reg and output is a mem.
15967 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15969 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15970 also for flag_dwarf2_cfi_asm.
15972 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15974 PR rtl-optimization/85342
15975 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15976 a bool scalar var inside of the loop instead. Don't try to update
15977 recog_data.operand after failed apply_change_group.
15979 2018-04-12 Tom de Vries <tom@codesourcery.com>
15982 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15983 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15984 array with flexible array member as array without given dimension.
15985 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15986 argument for undefined param to true.
15988 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15991 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15992 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15993 from PowerPC section.
15994 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15995 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15996 help text that is too long.
15997 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15998 help text that is too long.
15999 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
16000 help text that is too long.
16002 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
16004 * config/alpha/alpha.md (stack_probe_internal): Rename
16005 from "probe_stack". Update all callers.
16007 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16009 PR rtl-optimization/84566
16010 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
16011 sched_macro_fuse_insns.
16013 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16016 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
16017 (compute_block_dependences): ... from here.
16019 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16021 PR tree-optimization/85331
16022 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
16023 from int to HOST_WIDE_INT.
16025 2018-04-11 Martin Jambor <mjambor@suse.cz>
16028 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
16029 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
16030 not the same as the source val.
16031 (cgraph_edge_brings_value_p): New parameter.
16032 (gather_edges_for_value): Pass destination value to
16033 cgraph_edge_brings_value_p.
16034 (perhaps_add_new_callers): Likewise.
16035 (get_info_about_necessary_edges): Likewise and exclude values brought
16036 only by self-recursive edges.
16037 (create_specialized_node): Redirect only clones of self-calling edges.
16038 (+self_recursive_pass_through_p): New function.
16039 (find_more_scalar_values_for_callers_subset): Use it.
16040 (find_aggregate_values_for_callers_subset): Likewise.
16041 (known_aggs_to_agg_replacement_list): Removed.
16042 (decide_whether_version_node): Re-calculate known constants for all
16043 remaining context clones.
16045 2018-04-11 Richard Biener <rguenther@suse.de>
16048 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
16049 from early DWARF output.
16050 (dwarf2out_early_finish): Output line info unconditionally into
16051 early DWARF and add reference to it.
16053 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16056 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
16057 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
16058 other than V2DFmode using iptr mode attribute.
16059 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
16061 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16063 PR rtl-optimization/84659
16064 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
16066 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16069 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
16071 (output_loc_list): Pass address of a dummy size variable even in the
16072 locview handling loop.
16073 (index_location_lists): Add comment on why skip_loc_list_entry can't
16076 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
16079 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
16082 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
16085 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16086 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
16087 and -mstring-compare-inline-limit.
16089 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16092 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
16093 for stack clash protection in a register whenever we need it to be in
16096 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16098 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
16099 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
16101 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16104 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
16106 (mlong-double-): Ditto.
16107 * config/rs6000/sysv4.opt (msdata=): Ditto.
16108 (mtls-size=): Ditto.
16110 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16112 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16113 erroneous entries for
16114 "vector int vec_ldl (int, long int *)", and
16115 "vector unsigned int vec_ldl (int, unsigned long int *)".
16116 Add comments and entries for
16117 "vector bool char vec_ldl (int, bool char *)",
16118 "vector bool short vec_ldl (int, bool short *)",
16119 "vector bool int vec_ldl (int, bool int *)",
16120 "vector bool long long vec_ldl (int, bool long long *)",
16121 "vector pixel vec_ldl (int, pixel *)",
16122 "vector long long vec_ldl (int, long long *)",
16123 "vector unsigned long long vec_ldl (int, unsigned long long *)".
16124 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
16125 type tree bool_long_long_type_node and correct definition of
16126 bool_V2DI_type_node to make reference to this new type tree.
16127 (rs6000_mangle_type): Replace erroneous reference to
16128 bool_long_type_node with bool_long_long_type_node.
16129 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
16130 comments to emphasize sign distinctions for char and int types and
16131 replace RS6000_BTI_bool_long constant with
16132 RS6000_BTI_bool_long_long constant. Also add comment to restrict
16133 use of RS6000_BTI_pixel.
16134 (bool_long_type_node): Remove this macro definition.
16135 (bool_long_long_type_node): New macro definition
16137 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16139 PR rtl-optimization/85300
16140 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
16141 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
16142 simplify_unary_operation fails.
16144 2018-04-10 Martin Liska <mliska@suse.cz>
16146 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
16147 cgraph_edge and ipa_ref.
16149 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16153 * config/i386/sse.md
16154 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
16155 computation of the VEC_MERGE selector from mask.
16156 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
16157 Fix decoding of the VEC_MERGE selector into mask.
16159 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16161 PR tree-optimization/85286
16162 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16164 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16166 * final.c (final_1): Set insn_last_address as well as
16167 insn_current_address.
16169 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16172 * explow.c (emit_stack_probe): Call validize_mem on memory location
16173 before passing it to gen_probe_stack. Create address operand and
16174 legitimize it for the probe_stack_address case.
16176 2018-04-09 Jan Hubicka <jh@suse.cz>
16179 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
16180 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
16181 * tree.c (free_lang_data_in_type): Fix handling of binfos;
16183 (free_lang_data): Rebuild type inheritance graph.
16185 2018-04-09 Martin Sebor <msebor@redhat.com>
16187 * invoke.texi (-finline-small-functions): Mention other optimization
16189 (-findirect-inlining, -fpartial-inlining): Same.
16190 (-finline-functions-called-once): Same.
16191 (-freorder-blocks-and-partition): Same.
16193 2018-04-09 Jan Hubicka <jh@suse.cz>
16196 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
16197 jumps; choose last target that matches the criteria (i.e.
16198 no partition changes for non-crossing jumps).
16199 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
16200 support for redirecting crossing jumps to non-crossing.
16202 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
16204 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
16205 also for naked functions.
16207 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
16209 * config/arc/arc.md (add_shift): New pattern.
16210 (add_shift2): Likewise.
16211 (sub_shift): Likewise.
16212 (sub_shift_cmp0_noout): Likewise.
16213 (compare_si_ashiftsi): Likewise.
16214 (xbfu_cmp0_noout): New combine pattern.
16215 (xbfu_cmp0"): Likewise.
16216 (movsi_set_cc_insn): Place the predicable variant first.
16217 (commutative_binary_cmp0_noout): Remove clobber.
16218 (commutative_binary_cmp0): New pattern.
16219 (noncommutative_binary_cmp0): Likewise.
16220 (noncommutative_binary_cmp0_noout): Likewise.
16221 (noncommutative_binary_comparison_result_used): Removed.
16222 (rsub_cmp0): New pattern.
16223 (rsub_cmp0_noout): Likewise.
16224 (extzvsi): Changed, keep only meaningful variants.
16225 (SQH, SEZ): New iterators.
16226 (SQH_postfix): New mode attribute.
16227 (SEZ_prefix): New code attribute.
16228 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
16229 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
16230 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
16231 of numerical value.
16232 (noncommutative_operator): Check the availability of barrel
16235 2018-04-09 Richard Biener <rguenther@suse.de>
16237 PR tree-optimization/85284
16238 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
16239 Only use the niter constraining form of simple_iv when the exit
16240 is always executed.
16242 2018-04-09 Tom de Vries <tom@codesourcery.com>
16245 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16246 (define_expand "*memory_barrier"): New define_expand.
16247 (define_insn "memory_barrier"): New insn.
16249 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16251 PR rtl-optimization/80463
16252 PR rtl-optimization/83972
16253 PR rtl-optimization/83480
16255 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
16256 correct producer for the insn.
16257 (tidy_control_flow): Fixup seqnos in case of debug insns.
16259 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16261 PR rtl-optimization/83913
16263 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
16264 different sched-times when merging exprs.
16266 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16268 PR rtl-optimization/83962
16270 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
16271 tidy_fallthru_edge and tidy_control_flow.
16273 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16275 PR rtl-optimization/83530
16277 * sel-sched.c (force_next_insn): New global variable.
16278 (remove_insn_for_debug): When force_next_insn is true, also leave only
16279 next insn in the ready list.
16280 (sel_sched_region): When the region wasn't scheduled, make another pass
16281 over it with force_next_insn set to 1.
16283 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
16285 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
16287 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16288 for interrupt control.
16289 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16290 functions for interrupt control.
16291 * config/nds32/nds32-intrinsic.md: Likewise.
16292 * config/nds32/nds32_intrinsic.h: Likewise.
16293 * config/nds32/nds32.h (nds32_builtins): Likewise.
16295 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16297 * config/nds32/nds32.c (nds32_init_machine_status,
16298 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
16299 strict_aligned_p field.
16300 (nds32_expand_to_rtl_hook): New function.
16301 (TARGET_EXPAND_TO_RTL_HOOK): Define.
16302 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
16304 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16305 Chung-Ju Wu <jasonwucj@gmail.com>
16307 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
16308 * config/nds32/nds32-n7.md: New file.
16309 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
16312 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
16313 * config/nds32/nds32.md (pipeline_model): Add n7.
16314 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
16315 * config/nds32/pipelines.md: Include n7 settings.
16317 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16318 Chung-Ju Wu <jasonwucj@gmail.com>
16320 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
16321 * config/nds32/nds32-e8.md: New file.
16322 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
16323 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
16325 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
16326 * config/nds32/nds32.md (pipeline_model): Add e8.
16327 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
16328 * config/nds32/pipelines.md: Include e8 settings.
16330 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16331 Chung-Ju Wu <jasonwucj@gmail.com>
16333 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
16334 * config/nds32/nds32-n8.md: New file.
16335 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
16336 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
16338 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
16339 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
16340 * config/nds32/nds32.md (pipeline_model): Add n8.
16341 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
16342 * config/nds32/pipelines.md: Include n8 settings.
16344 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16345 Chung-Ju Wu <jasonwucj@gmail.com>
16347 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
16348 * config/nds32/nds32-n9-2r1w.md: New file.
16349 * config/nds32/nds32-n9-3r2w.md: New file.
16350 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
16351 nds32_register_ports): New or modify for cpu n9.
16352 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
16354 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
16355 * config/nds32/nds32-utils.c: New file.
16356 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
16357 TARGET_MUL_SLOW): Define.
16358 * config/nds32/nds32.md (pipeline_model): New attribute.
16359 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
16360 New options that support cpu n9.
16361 * config/nds32/pipelines.md: Include n9 settings.
16362 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
16364 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16366 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
16367 information if necessary.
16368 (output_cond_branch_compare_zero): Likewise.
16369 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
16370 (nds32_target_alignment): Refine for alignment.
16371 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
16372 (FUNCTION_BOUNDARY): Modify.
16373 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
16375 * config/nds32/nds32.opt (malways-align, malign-functions): New.
16377 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
16379 * config/nds32/constants.md (unspec_volatile_element): Add values for
16380 TLB operation and data prefetch.
16381 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16382 functions for TLB operation and data prefetch.
16383 * config/nds32/nds32-intrinsic.md: Likewise.
16384 * config/nds32/nds32_intrinsic.h: Likewise.
16385 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
16386 (nds32_print_operand): Likewise.
16387 * config/nds32/nds32.h (nds32_builtins): Likewise.
16389 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
16390 Andrew Pinski <pinsika@gcc.gnu.org>
16392 PR middle-end/82976
16393 * match.pd: Use constant_boolean_node of correct type instead of
16394 boolean_true_node or boolean_false_node for simplifying
16395 pointer comparisons to zero.
16397 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16399 PR tree-optimization/80021
16400 * tree.c (verify_type_variant): Make error call in verify_variant_match
16401 translatable and remove final full stop.
16403 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16405 * config/nds32/constants.md (unspec_volatile_element): Add
16406 UNSPEC_VOLATILE_EH_RETURN.
16407 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
16408 nds32_output_stack_pop): Support dwarf exception handling process.
16409 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
16410 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
16411 exception handling process.
16412 (nds32_compute_stack_frame): Likewise.
16413 (nds32_return_addr_rtx): Likewise.
16414 (nds32_initial_elimination_offset): Likewise.
16415 (nds32_expand_prologue): Likewise.
16416 (nds32_expand_epilogue): Likewise.
16417 (nds32_dynamic_chain_address): New function.
16418 * config/nds32/nds32.h (machine_function): Add fields for dwarf
16419 exception handling.
16420 (DYNAMIC_CHAIN_ADDRESS): Define.
16421 (EH_RETURN_DATA_REGNO): Define.
16422 (EH_RETURN_STACKADJ_RTX): Define.
16423 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
16424 patterns for dwarf exception handling.
16426 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16428 * config/nds32/nds32.h: Clean up obsolete macros.
16430 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16432 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16433 Add enum values for particular instructions.
16434 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16435 particular intrinsic functions.
16436 * config/nds32/nds32-intrinsic.md: Likewise.
16437 * config/nds32/nds32_intrinsic.h: Likewise.
16438 * config/nds32/nds32.h (nds32_builtins): Likewise.
16439 * config/nds32/nds32.md (type): Add pbsad and pbsada.
16440 (btst, ave): New patterns for particular instructions.
16442 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16444 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16445 Add enum values for atomic load/store and memory sync.
16446 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16448 * config/nds32/nds32-intrinsic.md: Likewise.
16449 * config/nds32/nds32_intrinsic.h: Likewise.
16450 * config/nds32/nds32.h (nds32_builtins): Likewise.
16452 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16454 PR tree-optimization/85257
16455 * fold-const.c (native_encode_vector): If not all elts could fit
16456 and off is -1, return 0 rather than offset.
16457 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16458 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16459 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
16460 adjust buffer in native_interpret_expr call.
16462 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16464 * config/nds32/constants.md (unspec_volatile_element): Add cache
16465 control enum values.
16466 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16467 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16468 * config/nds32/nds32.c (nds32_cctl_names): New.
16469 (nds32_print_operand): Handle cache control register names.
16470 * config/nds32/nds32.h (nds32_builtins): New enum values.
16471 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16473 * config/nds32/nds32.md (type): Add mmu.
16474 * config/nds32/pipelines.md (simple_insn): Add mmu.
16476 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16478 * config/nds32/nds32.md (type): Remove call.
16479 * config/nds32/pipelines.md (simple_insn): Likewise.
16481 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16483 * config/nds32/constants.md (unspec_volatile_element): Add
16484 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16485 UNSPEC_VOLATILE_FMFCFG.
16486 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16487 description for fmfcfg and fmfcsr.
16488 (bdesc_1arg): Add fmtcsr.
16489 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16490 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16491 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16492 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16493 unspec_fmfcfg): New patterns.
16494 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16495 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16496 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16497 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16498 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16499 __nds32__fmfcfg): Define.
16501 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16503 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16504 intrinsic register names.
16505 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16506 intrinsic register enum values and macros.
16508 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16510 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16511 for load/store addressing form.
16512 (nds32_print_operand_address): Likewise.
16514 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16517 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16518 based on LABEL_REF. Remove useless assertion.
16519 (pic_address_needs_scratch): Fix formatting.
16520 (sparc_legitimize_pic_address): Minor tweaks.
16521 (sparc_delegitimize_address): Adjust assertion accordingly.
16522 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16523 into symbolic_operand.
16524 (movsi_high_pic_label_ref): Likewise.
16525 (movsi_lo_sum_pic_label_ref): Likewise.
16526 (movdi_pic_label_ref): Likewise.
16527 (movdi_high_pic_label_ref): Likewise.
16528 (movdi_lo_sum_pic_label_ref): Likewise.
16530 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16532 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16533 custom LIB_SPEC setup.
16535 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16536 Kito Cheng <kito.cheng@gmail.com>
16538 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16539 * config/riscv/freebsd.h: New.
16541 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16543 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16544 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16547 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16548 Kito Cheng <kito.cheng@gmail.com>
16550 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16551 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16552 * config/nds32/nds32-protos.h (nds32_output_call,
16553 nds32_output_return): Declare.
16554 * config/nds32/nds32.md: Refine all the call and return patterns.
16556 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16559 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16560 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16562 PR rtl-optimization/84872
16563 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16564 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16565 EDGE_CROSSING edge.
16567 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16569 * expr.c (copy_blkmode_to_reg): Revert 254862.
16570 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16572 2018-04-06 Richard Biener <rguenther@suse.de>
16574 PR middle-end/85244
16575 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16576 after seeing a component reference with an adjacent field. Treat
16577 refs to arrays at struct end of external decls similar to
16578 refs to unconstrained commons.
16580 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16583 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16584 look through SAVE_EXPRs with non-side-effects argument. Adjust
16586 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16589 2018-04-06 Richard Biener <rguenther@suse.de>
16591 PR middle-end/85180
16592 * alias.c (find_base_term): New wrapper around find_base_term
16593 unwinding CSELIB_VAL_PTR changes.
16594 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16597 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16599 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16601 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16602 constant definitions.
16603 ("nop"): lr 0,0 -> nopr r0
16604 ("nop_lr0", "nop_lr1"): New insn definitions.
16606 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16608 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16609 NDS32_V3PUSH_AVAILABLE_P macro.
16611 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16612 Chung-Ju Wu <jasonwucj@gmail.com>
16614 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16615 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16616 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16617 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16618 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16619 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16620 * config/nds32/constraints.md: New constraints and checking for hard
16621 float configuration.
16622 * config/nds32/iterators.md: New mode iterator and attribute for hard
16623 float configuration.
16624 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16626 * config/nds32/nds32-fpu.md: New file.
16627 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16628 deal with hard float code generation.
16629 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16631 (abi_type, float_reg_number): New enum type.
16632 * config/nds32/nds32-predicates.c: New predicates for hard float.
16633 * config/nds32/nds32-protos.h: Declare functions for hard float.
16634 * config/nds32/nds32.c: Implementation for hard float configuration.
16635 * config/nds32/nds32.h: Definitions for hard float configuration.
16636 * config/nds32/nds32.md: Include hard float machine description and
16637 modify patterns for hard float configuration.
16638 * config/nds32/nds32.opt: New options for hard float configuration.
16639 * config/nds32/predicates.md: New predicates for hard float
16642 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16644 * common/config/nds32/nds32-common.c
16645 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16647 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16649 PR middle-end/85195
16650 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16651 CONSTRUCTOR_ELT (ctor, ...)->value.
16653 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16656 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16658 2018-04-05 Tom de Vries <tom@codesourcery.com>
16661 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16664 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16665 Kito Cheng <kito.cheng@gmail.com>
16667 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16668 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16669 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16670 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16672 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16673 Kito Cheng <kito.cheng@gmail.com>
16675 * config/nds32/constraints.md (Ufe): New memory constraint.
16676 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16677 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16678 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16680 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16681 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16683 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16685 * config/nds32/nds32.md: Use optimize_size in the condition for
16686 alu-shift instructions.
16688 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16690 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16692 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16694 * config/nds32/nds32.md (negsi2): Refine pattern.
16696 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16697 Chung-Ju Wu <jasonwucj@gmail.com>
16699 * config/nds32/iterators.md (shift_rotate): New code iterator.
16700 (shift): New code attribute.
16701 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16702 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16703 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16704 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16705 bit-wise operations.
16706 (andsi3, *andsi3): Ditto.
16707 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16708 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16709 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16710 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16711 nds32_ior_operand, nds32_xor_operand): New predicates.
16713 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16715 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16716 (addsi3, subsi3): ... this.
16718 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16720 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16722 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16724 * config/nds32/nds32.md: Adjust indention.
16726 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16728 * config/nds32/nds32.md (feature): New attribute.
16730 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16732 * config/nds32/nds32.md (subtype): New attribute.
16734 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
16737 * config/arm/arm-builtins.c (arm_expand_builtin): Change
16738 expansion to perform a bitwise AND of the argument followed by a
16739 boolean negation of the result.
16741 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
16743 PR rtl-optimization/84878
16744 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16745 the basic block. Assert the use reference is not artificial and that
16746 it has an associated insn.
16748 2018-04-04 Michael Matz <matz@suse.de>
16750 * builtins.c (compute_objsize): Pass correct operand
16751 to array_at_struct_end_p.
16753 2018-04-04 Richard Biener <rguenther@suse.de>
16756 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16757 from contexts for DINFO_LEVEL_TERSE and below.
16759 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16761 * config/nds32/nds32-doubleword.md (move_<mode>): Require
16762 resiter_operand condition.
16763 * config/nds32/nds32.md (*move<mode>): Ditto.
16765 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16766 Monk Chiang <sh.chiang04@gmail.com>
16768 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16770 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16772 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16774 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16775 Kito Cheng <kito.cheng@gmail.com>
16777 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16778 nds32_cond_code_str, output_cond_branch,
16779 output_cond_branch_compare_zero, nds32_expand_cbranch,
16780 nds32_expand_cstore, nds32_expand_movcc,
16781 nds32_output_cbranchsi4_equality_zero,
16782 nds32_output_cbranchsi4_equality_reg,
16783 nds32_output_cbranchsi4_equality_reg_or_const_int,
16784 nds32_output_cbranchsi4_greater_less_zero: New functions.
16785 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16786 nds32_expand_cstore, nds32_expand_movcc,
16787 nds32_output_cbranchsi4_equality_zero,
16788 nds32_output_cbranchsi4_equality_reg,
16789 nds32_output_cbranchsi4_equality_reg_or_const_int,
16790 nds32_output_cbranchsi4_greater_less_zero): Declare.
16791 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16792 nds32_rimm11s_operand): New predicates.
16793 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16794 * config/nds32/nds32.md: Rewrite all the branch and conditional move
16797 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16799 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16800 * config/nds32/nds32.md: Ditto.
16801 * config/nds32/pipelines.md: Ditto.
16803 2018-04-04 Richard Biener <rguenther@suse.de>
16805 PR tree-optimization/85168
16806 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16807 propagating abnormals.
16809 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16811 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16813 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16814 Kito Cheng <kito.cheng@gmail.com>
16816 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16817 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16818 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16819 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16820 * config/nds32/nds32.md (sibcall_internal): New.
16821 (sibcall_register): Remove.
16822 (sibcall_immediate): Remove.
16823 (sibcall_value_internal): New.
16824 (sibcall_value_register): Remove.
16825 (sibcall_value_immediate): Remove.
16826 * config/nds32/predicates.md (nds32_general_register_operand): New.
16827 (nds32_call_address_operand): New.
16829 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16831 PR rtl-optimization/85167
16832 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16833 bb_defs if *split_p, instead preinitialize it to NULL.
16835 PR tree-optimization/85156
16836 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16837 evaluating the argument multiple times.
16839 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16841 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16843 (_mm_cvtpd_ps): Likewise.
16844 (_mm_cvttpd_epi32): Likewise.
16845 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16846 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16847 vector, pixel, and bool following altivec.h include.
16849 2018-04-03 Martin Sebor <msebor@redhat.com>
16851 * doc/extend.texi (Common Function Attributes): Clarify.
16852 (const attribute): Likewise.
16853 (pure attribute): Likewise.
16855 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16858 * config/i386/i386.c (ix86_expand_vector_set): Use
16859 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16861 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16863 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16864 instructions when changing rounding bits to preserve precision bits
16865 in the x87 control word.
16867 2018-04-03 Martin Liska <mliska@suse.cz>
16869 PR tree-optimization/82491
16870 * rtl.h (strip_offset_and_add): Replace += suboffset with
16871 poly_uint64 () + suboffset.
16873 2018-03-29 Martin Liska <mliska@suse.cz>
16874 Martin Jambor <mjambor@suse.cz>
16877 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16878 param_type is not an integral or pointer type.
16880 2018-04-03 Richard Biener <rguenther@suse.de>
16882 * sese.h (recompute_all_dominators): Remove.
16884 2018-04-02 Martin Sebor <msebor@redhat.com>
16886 * doc/invoke.texi (-Wrestrict): Fix typos.
16888 2018-04-02 Jim Wilson <jimw@sifive.com>
16890 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16891 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16892 (<optab>di3, <optab>si3_extend): Likewise.
16893 (<optab>si3_mask, <optab>si3_mask_1): New.
16894 (<optab>di3_mask, <optab>di3_mask_1): New.
16895 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16896 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16897 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16899 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16901 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16904 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16906 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16907 (nds32_canonicalize_comparison): New function.
16909 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16910 Kito Cheng <kito.cheng@gmail.com>
16911 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16913 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16914 * config/nds32/constants.md (unspec_volatile_element): Add
16915 UNSPEC_VOLATILE_RELAX_GROUP.
16916 * config/nds32/nds32-relax-opt.c: New file.
16917 * config/nds32/nds32-predicates.c
16918 (nds32_symbol_load_store_p): New function.
16919 * config/nds32/nds32-protos.h
16920 (nds32_symbol_load_store_p): Declare function.
16921 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16922 * config/nds32/nds32.c
16923 (nds32_register_pass): New function to register pass.
16924 (nds32_register_passes): New function to register passes.
16925 * config/nds32/nds32.md (relax_group): New pattern.
16926 * config/nds32/nds32.opt (mrelax-hint): New option.
16927 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16929 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16931 * config/nds32/t-nds32: Modify files dependency.
16933 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16935 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16936 (PROFILE_HOOK): Define its implementation.
16938 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16940 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16941 type and 32-bit size.
16943 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16945 PR middle-end/85090
16946 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16947 (V_128_256): New mode iterator.
16948 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16949 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16950 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16952 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16953 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16955 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16958 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16959 NaN inputs correctly.
16961 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16964 * config/rs6000/vsx.md (??r): New mode attribute.
16965 (*vsx_mov<mode>_64bit): Use it.
16966 (*vsx_mov<mode>_32bit): Likewise.
16968 2018-03-30 Martin Sebor <msebor@redhat.com>
16970 PR tree-optimization/84818
16971 * builtins.c (check_access): Use warning_n.
16973 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16976 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16978 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16981 2018-03-30 Julia Koval <julia.koval@intel.com>
16984 * x86-tune.def (movx, partial_reg_dependency): Enable for
16987 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16989 PR inline-asm/84985
16990 * lra-constraints.c (process_alt_operands): Move setting
16991 this_alternative_matches below.
16993 2018-03-29 Martin Liska <mliska@suse.cz>
16996 * doc/invoke.texi: Document how LTO works with debug info.
16997 Describe auto-load support of binutils. Mention 'x86-64'
16998 as valid option value of -march option.
17000 2018-03-29 Jakub Jelinek <jakub@redhat.com>
17002 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
17005 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
17006 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
17007 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
17010 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17013 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
17014 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
17015 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17016 for RS6000_BTM_POWERPC64.
17017 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
17018 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
17019 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
17022 (DIVDEU): Likewise.
17024 2018-03-28 Carl Love <cel@us.ibm.com>
17027 2017-09-27 Carl Love <cel@us.ibm.com>
17029 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
17030 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
17031 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
17034 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17036 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
17037 instead of __vector bool.
17038 (_mm_max_pu8): Likewise.
17039 (_mm_min_pi16): Likewise.
17041 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17044 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
17045 (DIVWEUO): Likewise.
17046 (DIVDEO): Likewise.
17047 (DIVDEUO): Likewise.
17048 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
17049 DIVWEUO and DIVDEUO.
17050 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
17051 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
17052 (div_extend): Likewise.
17053 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
17055 (__builtin_divweuo): Likewise.
17056 (__builtin_divdeo): Likewise.
17057 (__builtin_divdeuo): Likewise.
17059 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17062 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
17063 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
17065 PR tree-optimization/82004
17066 * gimple-match-head.c (optimize_pow_to_exp): New function.
17067 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
17068 Don't fold to exp if optimize_pow_to_exp is false.
17070 2018-03-28 Martin Liska <mliska@suse.cz>
17073 * calls.c (initialize_argument_information): Fix trailing space.
17074 * common.opt: Fix typo and provide better explanation for
17075 -fsanitize-coverage option.
17076 * config/i386/i386.opt: Fix typo.
17078 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17079 Martin Liska <mliska@suse.cz>
17082 * gimplify.c (asan_poison_variable): Don't do the check for
17083 gimplify_omp_ctxp here.
17084 (gimplify_decl_expr): Do it here.
17085 (gimplify_target_expr): Likewise.
17087 2018-03-28 Martin Liska <mliska@suse.cz>
17090 * config/i386/i386.c (ix86_function_arg_advance): Do not call
17091 chkp_type_bounds_count if MPX is not enabled.
17093 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
17095 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
17097 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
17100 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
17101 function to create the function decl for complex long double
17102 multiply and divide for -mabi=ieeelongdouble.
17103 (init_float128_ieee): Call it.
17105 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
17108 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17109 -fcf-protection=branch -mibt.
17110 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
17112 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17115 * config/arm/arm.c (arm_valid_symbolic_address): Handle
17116 arm_word_relocations.
17118 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
17121 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
17122 extern array declarations.
17124 2018-03-27 Richard Biener <rguenther@suse.de>
17126 PR middle-end/84067
17127 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
17128 explicit single_use checks.
17130 2018-03-27 Richard Biener <rguenther@suse.de>
17132 PR tree-optimization/85082
17133 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
17136 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17138 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
17139 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
17140 Turn on fasynchronous-unwind-tables and funwind-tables.
17142 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
17145 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
17146 (*bmi_blsr_<mode>_ccz): Ditto.
17148 2018-03-26 Tom de Vries <tom@codesourcery.com>
17150 PR tree-optimization/85063
17151 * omp-general.c (offloading_function_p): New function. Factor out
17153 * omp-offload.c (pass_omp_target_link::gate): ... here.
17154 * omp-general.h (offloading_function_p): Declare.
17155 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
17156 with attribute omp declare target for offloading functions.
17158 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
17160 PR tree-optimization/84005
17161 * tree-data-ref.h (get_base_for_alignment): Declare.
17162 * tree-data-ref.c (get_base_for_alignment_1): New function.
17163 (get_base_for_alignment): Likewise.
17164 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
17165 get_base_for_alignment to find a suitable base object, instead
17166 of always using drb->base_address.
17168 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17170 PR inline-asm/85022
17171 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
17172 known size by default.
17174 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
17176 PR inline-asm/85030
17177 * lra-constraints.c (process_alt_operands): Don't match BLKmode
17178 and non BLKmode operands.
17180 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17183 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
17184 Clean up attributes.
17186 2018-03-23 Richard Biener <rguenther@suse.de>
17189 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
17190 we are going to emit early debug for LTO.
17192 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17194 PR inline-asm/85034
17195 * function.c (match_asm_constraints_1): Don't optimize if input
17196 doesn't satisfy general_operand predicate for output's mode.
17198 PR inline-asm/85022
17199 * alias.c (write_dependence_p): Don't require for x_canonicalized
17200 non-VOIDmode if x has VOIDmode.
17203 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
17204 just don't try to optimize it rather than assert it never happens.
17206 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17208 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
17209 macro expansions for definition of ST_INTERNAL_<mode> and
17210 LD_INTERNAL_<mode> builtins.
17211 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
17213 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
17215 (altivec_expand_st_builtin): Likewise.
17216 (altivec_expand_builtin): Remove calls to deleted functions.
17217 (rs6000_address_for_altivec): Delete this function.
17218 * config/rs6000/vector.md: Remove expands for
17219 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
17221 2018-03-22 Sudakshina Das <sudi.das@arm.com>
17224 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
17225 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
17226 re-computing once computed.
17227 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
17228 (arm_init_machine_status): Initialize
17229 machine->static_chain_stack_bytes.
17231 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17234 * doc/extend.texi: Add four new prototypes for vec_ld.
17235 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
17236 definitions for more logical presentation.
17237 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
17238 entries for V1TI variants of __builtin_altivec_ld builtin.
17239 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
17240 handling of V1TI variant of LVX icode pattern.
17241 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
17242 (rs6000_gimple_fold_builtin): Likewise.
17243 (altivec_init_builtins): Add code to define
17244 __builtin_altivec_lvx_v1ti function.
17246 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17248 PR inline-asm/84941
17249 * function.c (match_asm_constraints_1): Don't do the optimization
17250 if input isn't a REG, SUBREG, MEM or constant.
17252 2018-03-22 Tom de Vries <tom@codesourcery.com>
17254 PR tree-optimization/84956
17255 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
17256 bb_has_abnormal_pred.
17258 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17261 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17262 DECL_INITIAL (decl) to decl at the end.
17263 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
17264 adjust the comment.
17266 2018-03-21 Joseph Myers <joseph@codesourcery.com>
17268 * doc/extend.texi (__builtin_tgmath): Document when complex
17269 integer types are treated as _Complex _Float64.
17271 2018-03-21 Tom de Vries <tom@codesourcery.com>
17273 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
17275 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17277 PR tree-optimization/84960
17278 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
17279 if it is ENTRY block, move them into single succ of ENTRY in that case.
17281 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
17283 PR tree-optimization/84811
17284 * poly-int.h (poly_span_traits): Remove the T3 parameter and
17285 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
17286 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
17287 (known_subrange_p): Update accordingly. Cast each value involved
17288 in the size comparison, rather than casting the result of the
17291 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17293 PR tree-optimization/84982
17294 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
17295 by flipping the least significant bit rather than all bits from
17296 bitpos to bitpos + bitsize - 1.
17298 2018-03-21 Nathan Sidwell <nathan@acm.org>
17300 * doc/extend.texi (Deprecated Features): Remove mention of
17301 long-deleted deprecations.
17303 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17306 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17307 * configure: Regenerate.
17309 2018-03-21 Tom de Vries <tom@codesourcery.com>
17311 PR tree-optimization/83126
17312 * tree-parloops.c (num_phis): New function.
17313 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
17315 2018-03-21 Nathan Sidwell <nathan@acm.org>
17317 * doc/extend.texi (Deprecated Features): Update deprecated flags,
17318 mention anon-struct/union members and trailing attributes.
17320 2018-03-21 Bin Cheng <bin.cheng@arm.com>
17322 PR tree-optimization/84969
17323 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
17324 builtin memset partitions if they set different rhs values.
17326 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17328 PR rtl-optimization/84989
17329 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
17330 VEC_DUPLICATE with scalar result mode.
17332 2018-03-21 Martin Liska <mliska@suse.cz>
17335 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17336 not intended return statement.
17338 2018-03-21 Martin Liska <mliska@suse.cz>
17341 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
17342 (chkp_find_bound_slots_1): Limit number of iterations.
17344 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
17347 * Minor grammar fixes for x86 options.
17349 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17352 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
17353 holding REG_CFA_RESTORE notes, instead turn them into a USE.
17355 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
17358 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
17359 (altivec_lvx_<mode>_1op): Likewise.
17360 (altivec_stvx_<mode>_2op): Likewise.
17361 (altivec_stvx_<mode>_1op): Likewise.
17362 (altivec_lvx_<VM2:mode>): New define_expand.
17363 (altivec_stvx_<VM2:mode>): Likewise.
17364 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
17365 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17366 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
17367 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17368 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
17369 (rs6000_gen_lvx): Likewise.
17370 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
17371 (altivec_expand_stv_builtin): Likewise.
17372 (altivec_expand_builtin): Likewise.
17373 * config/rs6000/vector.md: Likewise.
17375 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17378 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
17381 2018-03-20 Richard Biener <rguenther@suse.de>
17384 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
17385 sign-conversions as zero, fall back to standard scalar_stmt
17388 2018-03-20 Martin Liska <mliska@suse.cz>
17391 * predict.c (rebuild_frequencies): Handle case when we have
17392 PROFILE_ABSENT, but flag_guess_branch_prob is false.
17394 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17397 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
17398 flag_section_anchors.
17399 * varasm.c (use_blocks_for_decl_p): Remove hack for
17400 dw2_force_const_mem.
17403 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
17405 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
17406 be created, use lowpart_subreg of operands[0] rather than operands[0]
17408 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
17409 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
17410 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
17411 and n constraint instead of aarch64_shift_imm_di and Usd.
17412 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
17413 (*aarch64_<optab>_reg_minus<mode>3): ... this.
17415 2018-03-20 Sudakshina Das <sudi.das@arm.com>
17418 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
17419 to favor GPR over NEON registers.
17420 (<shift>di3_neon): Likewise.
17422 2018-03-20 Tom de Vries <tom@codesourcery.com>
17425 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
17426 (nvptx_process_pars): Emit bar.sync asap and alap.
17428 2018-03-20 Tom de Vries <tom@codesourcery.com>
17431 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
17432 seen_label if seen_label is already set.
17434 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17437 * config/i386/i386.c (fold_builtin_cpu): For features above 31
17438 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
17439 Use 1U instead of 1. Formatting fixes.
17442 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17443 instead of TREE_TYPE (s1) for the return value.
17445 2018-03-19 Jakub Jelinek <jakub@redhat.com>
17447 PR tree-optimization/84946
17448 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17449 bitsize + bitsize in poly_uint64 rather than poly_int64.
17452 * dwarf2asm.c: Include fold-const.c.
17453 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17454 of decl rather than decl itself.
17456 PR rtl-optimization/84643
17457 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17459 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
17462 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17463 calling assemble_variable.
17465 2018-03-19 Sudakshina Das <sudi.das@arm.com>
17468 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17469 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17471 2018-03-19 Jim Wilson <jimw@sifive.com>
17474 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17475 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17476 (riscv_first_stack_step): Likewise.
17477 (riscv_option_override): Use STACK_BOUNDARY instead of
17478 MIN_STACK_BOUNDARY.
17479 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17480 MIN_STACK_BOUNDARY.
17481 (BIGGEST_ALIGNMENT): Set to 128.
17482 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17483 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17486 2018-03-19 Richard Biener <rguenther@suse.de>
17488 PR tree-optimization/84933
17489 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17490 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17492 2018-03-19 Richard Biener <rguenther@suse.de>
17494 PR tree-optimization/84859
17495 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17496 (cond_if_else_store_replacement): Perform sinking operation on
17497 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17498 Generalize what a BB with a single eligible store is.
17500 2018-03-19 Richard Biener <rguenther@suse.de>
17502 PR tree-optimization/84929
17503 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17504 chrec_is_positive against non-chrec arg.
17506 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17509 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17511 2018-03-18 Martin Liska <mliska@suse.cz>
17513 PR rtl-optimization/84635
17514 * regrename.c (build_def_use): Use matches_mode only when
17517 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17519 PR tree-optimization/84913
17520 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17521 vectorize chains of COND_EXPRs.
17523 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17525 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17527 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17529 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17531 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17533 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17535 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17536 Kito Cheng <kito.cheng@gmail.com>
17538 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17539 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17540 (nds32_adjust_reg_alloc_order): New function.
17541 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17543 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17545 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17546 nds32_print_operand, nds32_print_operand_address): Use
17547 HOST_WIDE_INT_PRINT_DEC instead.
17549 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17551 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17553 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17556 * config/i386/i386.c (initial_ix86_tune_features,
17557 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17558 unsigned long long.
17559 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17560 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17561 rather than 1u << ix86_tune. Formatting fix.
17562 (ix86_option_override_internal): Change ix86_arch_mask from
17563 unsigned int to unsigned HOST_WIDE_INT, initialize to
17564 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17565 (ix86_function_specific_restore): Likewise.
17567 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17570 * postreload.c (reload_combine_recognize_pattern): Perform
17571 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17572 truncate_int_for_mode the result for the destination's mode.
17575 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17576 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17578 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17581 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17584 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17585 regno_allocno_class_array and sorted_pseudos.
17586 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17587 insns where regno is used.
17589 2018-03-16 Martin Liska <mliska@suse.cz>
17592 * multiple_target.c (create_dispatcher_calls): Redirect
17593 reference in the symbol table.
17595 2018-03-16 Martin Liska <mliska@suse.cz>
17598 * multiple_target.c (create_dispatcher_calls): Redirect also
17601 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17605 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17606 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17607 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17609 2018-03-16 Julia Koval <julia.koval@intel.com>
17611 * doc/invoke.texi (Skylake Server): Add CLWB.
17612 Cannonlake): Remove CLWB.
17614 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17616 PR tree-optimization/84841
17617 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17619 (FLOAT_ONE_CONST_TYPE): Define.
17620 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17621 (sort_by_operand_rank): Put entries with higher constant_type last
17622 rather than first to match comments.
17624 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17626 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17629 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17632 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17633 check and use address of target always.
17635 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17638 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17639 (indirect_thunk_bnd_needed): Likewise.
17640 (indirect_thunks_used): Likewise.
17641 (indirect_thunks_bnd_used): Likewise.
17642 (indirect_return_needed): New.
17643 (indirect_return_bnd_needed): Likewise.
17644 (output_indirect_thunk_function): Add a bool argument for
17646 (output_indirect_thunk_function): Don't generate alias for
17647 function return thunk.
17648 (ix86_code_end): Call output_indirect_thunk_function to generate
17649 function return thunks.
17650 (ix86_output_function_return): Set indirect_return_bnd_needed
17651 and indirect_return_needed instead of indirect_thunk_bnd_needed
17652 and indirect_thunk_needed.
17654 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17656 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17657 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17658 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17660 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17661 Paul Hua <paul.hua.gm@gmail.com>
17664 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17666 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17668 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17669 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17670 resp. SFmode cases.
17672 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17675 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17676 instead of GET_MODE_SIZE when comparing Units.
17678 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17681 * varasm.c (hash_section): Return an unchangeble hash value
17682 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17683 Return !aarch64_can_use_per_function_literal_pools_p ().
17685 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17688 * optabs.c (emit_conditional_move): Pass address of cmode's copy
17689 rather than address of cmode as last argument to prepare_cmp_insn.
17691 2018-03-15 Julia Koval <julia.koval@intel.com>
17693 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17694 F_AVX512VNNI, F_AVX512BITALG): New.
17696 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
17699 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
17700 insn for floating-point loads and stores.
17702 2018-03-14 Carl Love <cel@us.ibm.com>
17704 * config/rs6000/rs6000-c.c: Add macro definitions for
17705 ALTIVEC_BUILTIN_VEC_PERMXOR.
17706 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17707 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17708 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17709 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17711 * config/doc/extend.texi: Add prototypes for vec_permxor.
17713 2018-03-14 David Malcolm <dmalcolm@redhat.com>
17716 * diagnostic-show-locus.c (class layout_point): Convert m_line
17717 from int to linenum_type.
17718 (line_span::comparator): Use linenum "compare" function when
17719 comparing line numbers.
17720 (test_line_span): New function.
17721 (layout_range::contains_point): Convert param "row" from int to
17723 (layout_range::intersects_line_p): Likewise.
17724 (layout::will_show_line_p): Likewise.
17725 (layout::print_source_line): Likewise.
17726 (layout::should_print_annotation_line_p): Likewise.
17727 (layout::print_annotation_line): Likewise.
17728 (layout::print_leading_fixits): Likewise.
17729 (layout::annotation_line_showed_range_p): Likewise.
17730 (struct line_corrections): Likewise for field m_row.
17731 (line_corrections::line_corrections): Likewise for param "row".
17732 (layout::print_trailing_fixits): Likewise.
17733 (layout::get_state_at_point): Likewise.
17734 (layout::get_x_bound_for_row): Likewise.
17735 (layout::print_line): Likewise.
17736 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17737 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17738 * input.c (selftest::test_linenum_comparisons): New function.
17739 (selftest::input_c_tests): Call it.
17740 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17741 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17742 * selftest.h (ASSERT_GT): New macro.
17743 (ASSERT_GT_AT): New macro.
17744 (ASSERT_LT): New macro.
17745 (ASSERT_LT_AT): New macro.
17747 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
17749 PR rtl-optimization/84780
17750 * combine.c (distribute_links): Don't make a link based on pc_rtx.
17752 2018-03-14 Martin Liska <mliska@suse.cz>
17754 * tree.c (record_node_allocation_statistics): Use
17755 get_stats_node_kind.
17756 (get_stats_node_kind): New function extracted from
17757 record_node_allocation_statistics.
17758 (free_node): Use get_stats_node_kind.
17760 2018-03-14 Richard Biener <rguenther@suse.de>
17762 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17763 that the value-set of ANTIC_IN doesn't grow.
17766 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17768 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17769 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17771 2018-03-14 Julia Koval <julia.koval@intel.com>
17773 * config.gcc (icelake-client, icelake-server): New.
17775 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17776 (initial_ix86_arch_features): Ditto.
17777 (PTA_SKYLAKE): Add SGX.
17778 (PTA_ICELAKE): Remove.
17779 (PTA_ICELAKE_CLIENT): New.
17780 (PTA_ICELAKE_SERVER): New.
17781 (ix86_option_override_internal): Split up icelake on icelake client and
17783 (get_builtin_code_for_version): Ditto.
17784 (fold_builtin_cpu): Ditto.
17785 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17786 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17787 * config/i386/i386.h (processor_type): Ditto.
17788 * doc/invoke.texi: Ditto.
17790 2018-03-14 Jakub Jelinek <jakub@redhat.com>
17793 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17794 INTEGER_CST offset, add it together with bitpos / 8 and
17795 sign extend based on POINTER_SIZE.
17799 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
17802 * config/i386/constraints.md (Yc): New register constraint.
17803 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17804 Use Yc constraint for alternative 2 of operand 0. Remove
17805 preferred_for_speed attribute.
17807 2018-03-14 Richard Biener <rguenther@suse.de>
17809 PR tree-optimization/84830
17810 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17811 with the old one to avoid oscillations.
17813 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17816 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17818 (assign_by_spills): Return a flag of reload assignment failure.
17819 Do not process the reload assignment failures. Do not spill other
17820 reload pseudos if they has the same reg class. Update n if
17822 (lra_assign): Add a return arg. Set up from the result of
17823 assign_by_spills call.
17824 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17825 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17826 usage_insns if it is not NULL.
17827 (spill_hard_reg_in_range): New function.
17828 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17829 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17830 function prototypes.
17831 (lra_assign): Change prototype.
17832 * lra.c (lra): Add code to deal with fails by splitting hard reg
17835 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17837 * config/riscv/riscv.opt (mrelax): New option.
17838 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17839 "norelax" when riscv_mrelax is disabled.
17840 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17842 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17845 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17846 reassociation for int modes.
17848 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17850 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17851 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17853 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17854 * config/aarch64/aarch64-sve.md
17855 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17856 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17857 (*extend<mode><Vwide>2): Rename to...
17858 (aarch64_sve_extend<mode><Vwide>2): ...this.
17859 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17860 renaming the old pattern to...
17861 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17863 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17864 define_expand, renaming the old pattern to...
17865 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17866 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17867 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17868 account when deciding which SVE instruction the optab should use.
17869 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17871 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17873 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17874 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17875 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17876 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17877 (tlsdesc_small_<mode>): Turn a define_expand and use
17878 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17879 (tlsdesc_small_advsimd_<mode>): ...this.
17880 (tlsdesc_small_sve_<mode>): New pattern.
17882 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17884 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17885 (UNSPEC_UMUL_HIGHPART): New constants.
17886 (MUL_HIGHPART): New int iteraor.
17887 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17888 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17890 (*<su>mul<mode>3_highpart): New define_insn.
17892 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17895 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17898 2018-03-13 Martin Liska <mliska@suse.cz>
17901 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17903 (sem_item_optimizer::~sem_item_optimizer): Release it.
17904 (sem_item_optimizer::merge_classes): Register variable aliases.
17905 (sem_item_optimizer::fixup_pt_set): New function.
17906 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17907 * ipa-icf.h: Declare new variables and functions.
17909 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17911 PR middle-end/84834
17912 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17913 integer_pow2p@2 and test integer_pow2p in condition.
17914 (A < 0 ? C : 0): Similarly for @1.
17916 PR middle-end/84831
17917 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17918 characters starting at p contain '\0' character, don't look beyond
17922 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17923 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17926 * reg-stack.c (change_stack): Change update_end var from int to
17927 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17928 also call set_block_for_insn on the newly added insns and rescan.
17931 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17932 on the last operand.
17935 * tree.c (stabilize_reference_1): Return save_expr (e) for
17936 STATEMENT_LIST even if it doesn't have side-effects.
17938 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17940 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17942 2018-03-12 Renlin Li <renlin.li@arm.com>
17944 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17945 aarch64_output_scalar_simd_mov_immediate.
17947 2018-03-12 Martin Sebor <msebor@redhat.com>
17949 PR tree-optimization/83456
17950 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17951 for perfectly overlapping calls to memcpy.
17952 (gimple_fold_builtin_memory_chk): Same.
17953 (gimple_fold_builtin_strcpy): Handle no-warning.
17954 (gimple_fold_builtin_stxcpy_chk): Same.
17955 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17957 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17959 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17960 parameter. Use it for SFmode.
17961 (rs6000_function_arg_advance_1): Adjust.
17962 (rs6000_function_arg): Adjust.
17963 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17965 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17967 PR rtl-optimization/84169
17968 PR rtl-optimization/84780
17969 * combine.c (can_combine_p): Check for a 2-insn combination whether
17970 the destination register is used between the two insns, too.
17972 2018-03-12 Richard Biener <rguenther@suse.de>
17974 PR tree-optimization/84803
17975 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17976 for refs DR analysis didn't process.
17978 2018-03-12 Richard Biener <rguenther@suse.de>
17980 PR tree-optimization/84777
17981 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17982 force-vectorize loops ignore whether we are optimizing for size.
17984 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17986 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17987 (TARGET_MD_ASM_ADJUST): Define.
17989 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17990 Kito Cheng <kito.cheng@gmail.com>
17991 Chung-Ju Wu <jasonwucj@gmail.com>
17993 * config/nds32/nds32.c (nds32_compute_stack_frame,
17994 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17995 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17996 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17997 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17998 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17999 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
18000 * config/nds32/nds32.md (prologue, epilogue): Use macro
18001 NDS32_V3PUSH_AVAILABLE_P to do checking.
18003 2018-03-11 Jakub Jelinek <jakub@redhat.com>
18006 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
18007 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
18008 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
18009 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
18010 addition of most attributes on !orig_type_die or the attribute not
18011 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
18013 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18014 Chung-Ju Wu <jasonwucj@gmail.com>
18016 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
18017 __NDS32_VH__ macro.
18018 * config/nds32/nds32.opt (mvh): New option.
18020 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18021 Chung-Ju Wu <jasonwucj@gmail.com>
18023 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
18025 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
18026 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
18029 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18030 Chung-Ju Wu <jasonwucj@gmail.com>
18032 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
18034 * config/nds32/nds32-multiple.md (strlensi): New pattern.
18035 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
18037 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
18038 Kito Cheng <kito.cheng@gmail.com>
18039 Chung-Ju Wu <jasonwucj@gmail.com>
18041 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
18042 UNSPEC_FFMISM and UNSPEC_FLMISM.
18043 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
18044 for ffb, ffmism and flmism.
18045 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
18046 (unspec_ffmism): Ditto.
18047 (unspec_flmism): Ditto.
18048 (nds32_expand_builtin_impl): Check if string extension is available.
18049 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
18050 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
18052 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
18055 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18058 * lra-assigns.c (assign_by_spills): Return a flag of reload
18059 assignment failure. Do not process the reload assignment
18060 failures. Do not spill other reload pseudos if they has the same
18062 (lra_assign): Add a return arg. Set up from the result of
18063 assign_by_spills call.
18064 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18065 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18066 usage_insns if it is not NULL.
18067 (spill_hard_reg_in_range): New function.
18068 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18069 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18070 function prototypes.
18071 (lra_assign): Change prototype.
18072 * lra.c (lra): Add code to deal with fails by splitting hard reg
18075 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
18078 * config/i386/i386.opt: Replace Enforcment with Enforcement.
18080 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
18083 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
18084 (dw_val_node): Add val_symbolic_view.
18085 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
18086 (symview_upper_bound): New.
18087 (new_line_info_table): Initialize symviews_since_reset.
18088 (dwarf2out_source_line): Count symviews_since_reset and set
18089 symview_upper_bound.
18090 (dw_val_equal_p): Handle symview.
18091 (add_AT_symview): New.
18092 (print_dw_val): Handle symview.
18093 (attr_checksum, attr_checksum_ordered): Likewise.
18094 (same_dw_val_p, size_of_die): Likewise.
18095 (value_format, output_die): Likewise.
18096 (add_high_low_attributes): Use add_AT_symview for entry_view.
18097 (dwarf2out_finish): Reset symview_upper_bound, clear
18100 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
18103 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
18104 Add strict argument and use it.
18105 (rs6000_split_multireg_move): Update for new strict argument.
18106 (mem_operand_gpr): Disallow all non-offsettable addresses.
18107 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
18109 2018-03-09 Jakub Jelinek <jakub@redhat.com>
18112 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
18113 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
18114 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18117 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
18118 decl, use remap_type if we want to use the type.
18120 2018-03-09 Martin Sebor <msebor@redhat.com>
18122 PR tree-optimization/84526
18123 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18125 (builtin_access::generic_overlap): Be prepared to handle non-array
18128 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
18130 PR rtl-optimization/84682
18131 * lra-constraints.c (process_address_1): Check is_address flag
18132 for address constraints.
18133 (process_alt_operands): Likewise.
18134 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
18135 preprocess_constraints.
18136 * recog.h (preprocess_constraints): Add oploc parameter.
18138 * recog.c (preprocess_constraints): Test address_operand for
18139 CT_ADDRESS constraints.
18141 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18144 * lra-assigns.c (assign_by_spills): Return a flag of reload
18145 assignment failure. Do not process the reload assignment
18146 failures. Do not spill other reload pseudos if they has the same
18148 (lra_assign): Add a return arg. Set up from the result of
18149 assign_by_spills call.
18150 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18151 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18152 usage_insns if it is not NULL.
18153 (spill_hard_reg_in_range): New function.
18154 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18155 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18156 function prototypes.
18157 (lra_assign): Change prototype.
18158 * lra.c (lra): Add code to deal with fails by splitting hard reg
18161 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18164 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
18165 Accept complain bool parameter. Only emit errors if it is true.
18166 (arm_parse_cpu_option_name): Likewise.
18167 (arm_target_thumb_only): Adjust callers of the above.
18168 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
18169 prototype to take a default true bool parameter.
18170 (arm_parse_arch_option_name): Likewise.
18172 2018-03-09 David Malcolm <dmalcolm@redhat.com>
18173 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
18177 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
18178 * configure: Regenerate.
18179 * configure.ac ("linker --version-script option"): New.
18180 ("linker soname option"): New.
18182 2018-03-09 Richard Biener <rguenther@suse.de>
18184 PR tree-optimization/84775
18185 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
18186 immediate uses of predicate stmts and mark them modified.
18189 PR tree-optimization/84178
18190 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18192 (version_loop_for_if_conversion): Delay update_ssa call.
18193 (tree_if_conversion): Delay update_ssa until after predicate
18196 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
18199 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
18200 when the function accesses prior frames.
18202 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18205 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
18206 gen_llsym, otherwise call maybe_gen_llsym.
18208 PR inline-asm/84742
18209 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
18210 has ',' character inside of it.
18212 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18215 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
18216 as clobbering CC_REGNUM.
18218 2018-03-08 Richard Biener <rguenther@suse.de>
18220 PR middle-end/84552
18221 * tree-scalar-evolution.c: Include tree-into-ssa.h.
18222 (follow_copies_to_constant): Do not follow SSA names registered
18225 2018-03-08 Richard Biener <rguenther@suse.de>
18227 PR tree-optimization/84178
18228 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18230 (version_loop_for_if_conversion): Delay update_ssa call.
18231 (tree_if_conversion): Delay update_ssa until after predicate
18234 2018-03-08 David Malcolm <dmalcolm@redhat.com>
18236 PR tree-optimization/84178
18237 * tree-if-conv.c (release_bb_predicate): Remove the
18238 the assertion that the stmts have NULL use_ops.
18239 Discard the statements, asserting that they haven't
18240 yet been added to a BB.
18242 2018-03-08 Richard Biener <rguenther@suse.de>
18244 PR tree-optimization/84746
18245 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
18246 (phi_translate): Pass in destination ANTIC_OUT set.
18247 (phi_translate_1): Likewise. For a simplified result lookup
18248 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
18249 (phi_translate_set): Adjust.
18250 (do_pre_regular_insertion): Likewise.
18251 (do_pre_partial_partial_insertion): Likewise.
18253 2018-03-08 Martin Liska <mliska@suse.cz>
18255 PR gcov-profile/84735
18256 * doc/gcov.texi: Document usage of profile files.
18257 * gcov-io.h: Document changes in the format.
18259 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
18263 * dwarf2out.c (struct dw_line_info_table): Update comments for
18265 (FORCE_RESET_NEXT_VIEW): New.
18266 (FORCE_RESETTING_VIEW_P): New.
18267 (RESETTING_VIEW_P): Check for -1 too.
18268 (ZERO_VIEW_P): Likewise.
18269 (new_line_info_table): Force-reset next view.
18270 (dwarf2out_begin_function): Likewise.
18271 (dwarf2out_source_line): Simplify zero_view_p initialization.
18272 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
18273 view directly. Omit view when omitting .loc at line 0.
18275 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18277 PR tree-optimization/84740
18278 * tree-switch-conversion.c (process_switch): Call build_constructors
18279 only if info.phi_count is non-zero.
18281 PR tree-optimization/84739
18282 * tree-tailcall.c (find_tail_calls): Check call arguments against
18283 DECL_ARGUMENTS (current_function_decl) rather than
18284 DECL_ARGUMENTS (func) when checking for tail recursion.
18286 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18288 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
18289 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
18290 Volker Reichelt's entry and add entries for people that perform
18291 GCC fuzzy testing and report numerous bugs.
18293 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
18296 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
18297 readonly data in sdata, if that is disabled.
18298 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
18299 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
18300 -mreadonly-in-sdata option.
18302 2018-03-07 Martin Sebor <msebor@redhat.com>
18304 PR tree-optimization/84468
18305 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
18306 basic block when looking for nul assignment.
18308 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
18311 * except.h (output_function_exception_table): Adjust prototype.
18312 * except.c (output_function_exception_table): Remove FNNAME parameter
18313 and add SECTION parameter. Ouput one part of the table at a time.
18314 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
18315 the first part of the exception table and emit unwind directives.
18316 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
18317 (i386_pe_seh_cold_init): Likewise.
18318 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
18319 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
18320 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
18321 (ix86_output_call_insn): Emit a nop in one more case for SEH.
18322 * config/i386/winnt.c: Include except.h.
18323 (struct seh_frame_state): Add reg_offset, after_prologue and
18324 in_cold_section fields.
18325 (i386_pe_seh_end_prologue): Set seh->after_prologue.
18326 (i386_pe_seh_cold_init): New function.
18327 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
18328 to seh->in_cold_section.
18329 (seh_emit_push): Record the offset of the push.
18330 (seh_emit_save): Record the offet of the save.
18331 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
18332 Test seh->after_prologue to disregard the epilogue.
18333 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
18334 (i386_pe_end_cold_function): New function.
18336 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18339 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
18340 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
18343 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
18345 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
18346 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
18348 PR middle-end/84723
18349 * multiple_target.c: Include tree-inline.h and intl.h.
18350 (expand_target_clones): Diagnose and fail if node->definition and
18351 !tree_versionable_function_p (node->decl).
18353 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
18355 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18357 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
18358 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
18359 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
18361 2018-03-06 Jakub Jelinek <jakub@redhat.com>
18364 * combine.c (try_combine): Use reg_or_subregno instead of handling
18365 just paradoxical SUBREGs and REGs.
18367 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
18369 * config/arc/arc.c (arc_finalize_pic): Remove function.
18370 (arc_must_save_register): We use single base PIC register, remove
18371 checks to save/restore the PIC register.
18372 (arc_expand_prologue): Likewise.
18373 * config/arc/arc-protos.h (arc_set_default_type_attributes):
18375 (arc_verify_short): Likewise.
18376 (arc_attr_type): Likewise.
18377 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
18378 (walk_stores): Likewise.
18379 (arc_address_cost): Make it static.
18380 (arc_verify_short): Likewise.
18381 (branch_dest): Likewise.
18382 (arc_attr_type): Likewise.
18383 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
18384 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
18385 (arc_final_prescan_insn): Remove inserting the nops due to
18386 hardware hazards. It is done in reorg step.
18387 (insn_length_variant_t): Remove.
18388 (insn_length_parameters_t): Likewise.
18389 (arc_insn_length_parameters): Likewise.
18390 (arc_get_insn_variants): Likewise.
18391 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
18393 2018-03-06 Jakub Jelinek <jakub@redhat.com>
18395 PR inline-asm/84683
18396 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
18399 PR tree-optimization/84687
18400 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
18402 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
18404 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18406 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
18407 Rename to ppc_speculation_barrier.
18408 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
18409 __builtin_ppc_speculation_barrier.
18411 2018-03-05 Jakub Jelinek <jakub@redhat.com>
18414 * combine.c (combine_simplify_rtx): Don't try to simplify if
18415 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
18418 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
18420 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
18421 to 32 bytes when compiling for POWER9.
18423 2018-03-05 Jakub Jelinek <jakub@redhat.com>
18426 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
18427 regparm >= 3 with no arg reg available also for calls with
18428 flag_force_indirect_call. Pass decl to ix86_function_regparm.
18431 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
18433 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
18435 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
18438 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
18440 2018-03-05 Richard Biener <rguenther@suse.de>
18442 PR tree-optimization/84486
18443 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18444 When inserting a __builtin_assume_aligned call set the LHS
18445 SSA name alignment info accordingly.
18447 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
18449 PR tree-optimization/84114
18450 * config/aarch64/aarch64.c (aarch64_reassociation_width)
18451 Avoid reassociation of FLOAT_MODE addition.
18453 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
18455 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18456 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18457 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18458 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18459 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18460 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18461 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18463 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18464 __builtin_ia32_wbinvd): New builtins.
18465 (SPECIAL_ARGS2): New.
18466 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18467 (SPECIAL_ARGS2): New.
18468 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18469 (ix86_valid_target_attribute_inner_p): Ditto.
18470 (ix86_init_mmx_sse_builtins): Add special_args2.
18471 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18472 TARGET_WBNOINVD_P): New.
18473 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18474 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18475 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18476 * config/i386/immintrin.h (_wbinvd): New intrinsic.
18477 * config/i386/pconfigintrin.h: New file.
18478 * config/i386/wbnoinvdintrin.h: Ditto.
18479 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18481 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18483 2018-03-05 Richard Biener <rguenther@suse.de>
18485 PR tree-optimization/84670
18486 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18488 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18489 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18490 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18491 doesn't grow if all successors (recursively) were visited at least
18494 2018-03-05 Richard Biener <rguenther@suse.de>
18496 PR tree-optimization/84650
18497 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18498 if executed in the loop pipeline.
18500 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18502 * doc/configfiles.texi (Configuration Files): Move info about
18503 conditionalizing $target-protos.h to...
18504 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18505 differs from $target-protos.h.
18507 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18508 Chung-Ju Wu <jasonwucj@gmail.com>
18510 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18511 * config/nds32/nds32-multiple.md (setmemsi): Define.
18512 * config/nds32/nds32-memory-manipulation.c
18513 (nds32_gen_dup_4_byte_to_word_value): New.
18514 (emit_setmem_word_loop): New.
18515 (emit_setmem_byte_loop): New.
18516 (nds32_expand_setmem_loop): New.
18517 (nds32_expand_setmem_loop_v3m): New.
18518 (nds32_expand_setmem_unroll): New.
18519 (nds32_expand_setmem): New.
18521 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18522 Chung-Ju Wu <jasonwucj@gmail.com>
18524 * config/nds32/nds32-memory-manipulation.c
18525 (nds32_emit_load_store): New.
18526 (nds32_emit_post_inc_load_store): New.
18527 (nds32_emit_mem_move): New.
18528 (nds32_emit_mem_move_block): New.
18529 (nds32_expand_movmemsi_loop_unknown_size): New.
18530 (nds32_expand_movmemsi_loop_known_size): New.
18531 (nds32_expand_movmemsi_loop): New.
18532 (nds32_expand_movmemsi_unroll): New.
18533 (nds32_expand_movmemqi): Rename ...
18534 (nds32_expand_movmemsi): ... to this.
18535 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18536 (movmemsi): ... to this.
18537 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18538 (nds32_expand_movmemsi): ... to this.
18540 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18541 Monk Chiang <sh.chiang04@gmail.com>
18542 Chung-Ju Wu <jasonwucj@gmail.com>
18544 * config/nds32/nds32-protos.h
18545 (nds32_expand_load_multiple): New arguments.
18546 (nds32_expand_store_multiple): Ditto.
18547 (nds32_valid_multiple_load_store): Rename ...
18548 (nds32_valid_multiple_load_store_p): ... to this.
18549 * config/nds32/nds32-memory-manipulation.c
18550 (nds32_expand_load_multiple): Refine implementation.
18551 (nds32_expand_store_multiple): Ditto.
18552 * config/nds32/nds32-multiple.md
18553 (load_multiple): Update nds32_expand_load_multiple interface.
18554 (store_multiple): Update nds32_expand_store_multiple interface.
18555 * config/nds32/nds32-predicates.c
18556 (nds32_valid_multiple_load_store): Rename ...
18557 (nds32_valid_multiple_load_store_p): ... to this and refine
18559 * config/nds32/predicates.md
18560 (nds32_load_multiple_and_update_address_operation): New predicate.
18561 (nds32_store_multiple_and_update_address_operation): New predicate.
18563 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18564 Chung-Ju Wu <jasonwucj@gmail.com>
18566 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18567 (combo): New attribute.
18568 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18570 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18572 * config/nds32/nds32.opt: Change -mcmodel= default value.
18574 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18575 Monk Chiang <sh.chiang04@gmail.com>
18576 Chung-Ju Wu <jasonwucj@gmail.com>
18578 * config/nds32/constants.md (unspec_element): New enum.
18579 * config/nds32/constraints.md (Umw): New constraint.
18580 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18581 * config/nds32/nds32-intrinsic.md: Likewise.
18582 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18583 (nds32_valid_smw_lwm_base_p): New.
18584 (nds32_output_smw_single_word): New.
18585 (nds32_output_lmw_single_word): New.
18586 (nds32_expand_unaligned_load): New.
18587 (nds32_expand_unaligned_store): New.
18588 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18589 (nds32_output_smw_single_word): Declare.
18590 (nds32_output_lmw_single_word): Declare.
18591 (nds32_expand_unaligned_load): Declare.
18592 (nds32_expand_unaligned_store): Declare.
18593 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18594 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18595 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18596 NDS32_BUILTIN_UASTORE_DW.
18597 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18600 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18601 Kito Cheng <kito.cheng@gmail.com>
18602 Chung-Ju Wu <jasonwucj@gmail.com>
18604 * config/nds32/nds32-intrinsic.c
18605 (nds32_expand_builtin_null_ftype_reg): Delete.
18606 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18607 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18608 (nds32_read_argument): New.
18609 (nds32_legitimize_target): Ditto.
18610 (nds32_legitimize_argument): Ditto.
18611 (nds32_check_constant_argument): Ditto.
18612 (nds32_expand_unop_builtin): Ditto.
18613 (nds32_expand_unopimm_builtin): Ditto.
18614 (nds32_expand_binop_builtin): Ditto.
18615 (nds32_builtin_decl_impl): Ditto.
18616 (builtin_description): Ditto.
18617 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18618 (nds32_init_builtins_impl): Ditto.
18619 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18620 (nds32_builtin_decl): New.
18621 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18622 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18624 2018-03-02 Jeff Law <law@redhat.com>
18626 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18627 (redundant_insn, fill_simple_delay_slots): Likewise.
18628 (fill_slots_from_thread): Likewise.
18629 * resource.c (mark_referenced_resources): Likewise.
18630 (mark_set_resources, find_dead_or_set_registers): Likewise.
18632 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18634 * substring-locations.h (format_warning_va): Formatting fix for
18635 ATTRIBUTE_GCC_DIAG.
18636 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18638 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18639 * substring-locations.c: Include intl.h.
18640 (format_warning_va): Turned into small wrapper around
18641 format_warning_n_va, renamed to ...
18642 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18643 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18645 (format_warning_at_substring_n): New function.
18646 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18647 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18648 format_warning_at_substring with just a shorter name instead of
18649 const function pointer.
18650 (fmtwarn_n): New function.
18651 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18652 appropriate, get rid of all the fmtstr temporaries, move conditionals
18653 with G_() wrapped string literals directly into fmtwarn arguments,
18654 cast dir.len to (int), formatting fixes.
18656 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18658 * doc/invoke.texi: Remove "Cilk Plus" references.
18660 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18661 Richard Biener <rguenther@suse.de>
18664 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18665 for error or warning attributes if CALL_FROM_THUNK_P is set.
18668 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18671 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18672 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18675 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18676 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18678 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18679 instead of -1U in last predictors element's probability member.
18681 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18684 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18685 arguments if they are comparable.
18687 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18689 PR tree-optimization/84634
18690 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18691 masks and masked_loop_p with a single loop_masks, making sure it's
18692 null for bb vectorization.
18694 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18696 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18697 (vect_analyze_data_ref_access): Use loop->safe_len rather than
18698 loop->force_vectorize to check whether there is no alias.
18700 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18703 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18705 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18707 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18708 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18709 instead of a loop around prev_real_insn.
18710 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18713 PR inline-asm/84625
18714 * config/i386/i386.c (ix86_print_operand): Use conditional
18715 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18718 2018-03-02 Richard Biener <rguenther@suse.de>
18720 PR tree-optimization/84427
18721 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18722 (bitmap_set_subtract_values): Rewrite to handle multiple
18725 (prune_clobbered_mems): Likewise.
18726 (phi_translate): Take edge instead of pred/phiblock.
18727 (phi_translate_1): Likewise.
18728 (phi_translate_set): Likewise. Insert all translated
18729 exprs for a value into the set, keeping possibly multiple
18730 expressions per value.
18731 (compute_antic_aux): Adjust for phi_translate changes.
18732 When intersecting union the expressions and prune those
18733 not in the final value set, keeping possibly multiple
18734 expressions per value. Do not use value-insertion
18735 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18736 all expressions. Add verification that the value-sets
18737 only shrink during iteration.
18738 (compute_partial_antic_aux): Adjust for the phi_translate changes.
18739 (do_pre_regular_insertion): Likewise.
18740 (do_pre_partial_partial_insertion): Likewise.
18742 2018-03-02 Richard Biener <rguenther@suse.de>
18745 * config/darwin.c (saved_debug_info_level): New static global.
18746 (darwin_asm_lto_start): Disable debug info generation for LTO out.
18747 (darwin_asm_lto_end): Restore debug info generation settings.
18749 2018-03-01 Martin Liska <mliska@suse.cz>
18752 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18753 volatile arguments.
18755 2018-03-01 Richard Biener <rguenther@suse.de>
18758 * dwarf2out.c (gen_variable_die): Properly handle late VLA
18759 type annotation with LTO when debug was disabled at compile-time.
18761 2018-03-01 Matthew Fortune <mfortune@gmail.com>
18763 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18765 (mips_final_postscan_insn): Likewise.
18767 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
18769 PR rtl-optimization/84528
18770 * alias.c (init_alias_target): Add commentary.
18771 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18772 a unique base value if the frame pointer is not eliminated
18773 to the stack pointer.
18775 2018-03-01 Tom de Vries <tom@codesourcery.com>
18777 PR rtl-optimization/83327
18778 * lra-int.h (hard_regs_spilled_into): Declare.
18779 * lra.c (hard_regs_spilled_into): Define.
18780 (init_reg_info): Init hard_regs_spilled_into.
18781 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18782 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18783 (process_bb_lives): Handle hard_regs_spilled_into.
18784 (lra_create_live_ranges_1): Before doing liveness propagation, clear
18785 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18787 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
18789 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18790 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18791 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18792 * config/rs6000/aix72.h: New file.
18794 2018-02-28 Jakub Jelinek <jakub@redhat.com>
18796 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18797 instead of warning_at with conditional singular and plural messages
18801 * stor-layout.c (update_alignment_for_field): For
18802 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18803 && !DECL_PACKED (field), do the alignment update, just use
18804 only desired_align instead of MAX (type_align, desired_align)
18806 (place_field): Don't do known_align < desired_align handling
18807 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18808 is non-NULL, instead do it after rli->prev_field handling and
18809 only if not within a bitfield word. For DECL_PACKED (field)
18810 use type_align of BITS_PER_UNIT.
18812 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18814 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18815 superfluous parentheses and trailing spaces.
18817 2018-02-28 Richard Biener <rguenther@suse.de>
18819 PR tree-optimization/84584
18820 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18821 SCoPs with fake exit edge.
18823 2018-02-28 Martin Liska <mliska@suse.cz>
18826 * timevar.c (timer::print): Fix format to properly print 100%
18829 2018-02-28 Richard Biener <rguenther@suse.de>
18831 PR middle-end/84607
18832 * genmatch.c (capture_info::walk_match): Do not mark
18833 captured expressions without operands as expr_p given
18834 they act more like predicates and should be subject to
18835 "lost tail" side-effect preserving.
18837 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18839 PR rtl-optimization/81611
18840 * auto-inc-dec.c (attempt_change): Move dead note from
18841 mem_insn if it's the next use of regno
18842 (find_address): Take address use of reg holding
18843 non-incremented value. Add parm to limit search to the named
18845 (merge_in_block): Attempt to use a mem insn that is the next
18846 use of the original regno.
18848 2018-02-27 Martin Sebor <msebor@redhat.com>
18851 * doc/invoke.texi (-Wmissing-attributes): New option.
18852 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18854 2018-02-27 Martin Sebor <msebor@redhat.com>
18856 PR translation/84207
18857 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18858 n argument to unsigned HOST_WIDE_INT.
18859 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18860 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18861 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18862 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18864 2018-02-27 Richard Biener <rguenther@suse.de>
18866 PR tree-optimization/84512
18867 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18868 Do not use the estimate returned from record_stmt_cost for
18869 the scalar iteration cost but sum properly using add_stmt_cost.
18871 2018-02-27 Richard Biener <rguenther@suse.de>
18873 PR tree-optimization/84466
18874 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18875 Adjust last change to less strictly validate use operands.
18877 2018-02-27 Martin Liska <mliska@suse.cz>
18879 PR gcov-profile/84548
18880 * gcov.c (process_file): Allow partial overlap and consider it
18881 also as group functions.
18882 (output_lines): Properly calculate range of lines for a group.
18884 2018-02-27 Martin Liska <mliska@suse.cz>
18886 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18887 'ggc' suffixes. Change first column width.
18888 (timer::print): Fix formatting of the column.
18890 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18892 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18893 preserve inline entry blocks for the sake of debug inline
18894 entry point markers alone.
18895 (remove_unused_locals): Suggest in comments a better place to
18896 force the preservation of inline entry blocks that are
18897 otherwise unused, but do not preserve them.
18899 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18901 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18903 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18906 * config/i386/constraints.md (Bs): Replace
18907 ix86_indirect_branch_register with
18908 TARGET_INDIRECT_BRANCH_REGISTER.
18910 * config/i386/i386.md (indirect_jump): Likewise.
18911 (tablejump): Likewise.
18912 (*sibcall_memory): Likewise.
18913 (*sibcall_value_memory): Likewise.
18914 Peepholes of indirect call and jump via memory: Likewise.
18915 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18916 (*sibcall_value_GOT_32): Likewise.
18917 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18918 (GOT_memory_operand): Likewise.
18919 (call_insn_operand): Likewise.
18920 (sibcall_insn_operand): Likewise.
18921 (GOT32_symbol_operand): Likewise.
18922 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18924 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18926 PR rtl-optimization/83496
18927 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18928 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18929 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18930 redundant insn, if any.
18931 (relax_delay_slots): Likewise.
18932 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18934 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18936 PR tree-optimization/83965
18937 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18938 that grouped statements are part of a reduction chain. Return
18939 true if the statement is not marked as a reduction itself but
18940 is part of a group.
18941 (vect_recog_dot_prod_pattern): Don't check whether the statement
18942 is part of a group here.
18943 (vect_recog_sad_pattern): Likewise.
18944 (vect_recog_widen_sum_pattern): Likewise.
18946 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18949 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18951 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18954 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18956 (ix86_output_indirect_function_return): New prototype.
18957 (ix86_split_simple_return_pop_internal): Likewise.
18958 * config/i386/i386.c (indirect_return_via_cx): New.
18959 (indirect_return_via_cx_bnd): Likewise.
18960 (indirect_thunk_name): Handle return va CX_REG.
18961 (output_indirect_thunk_function): Create alias for
18962 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18963 (ix86_output_indirect_jmp): Remove the bool argument.
18964 (ix86_output_indirect_function_return): New function.
18965 (ix86_split_simple_return_pop_internal): Likewise.
18966 * config/i386/i386.md (*indirect_jump): Don't pass false
18967 to ix86_output_indirect_jmp.
18968 (*tablejump_1): Likewise.
18969 (simple_return_pop_internal): Change it to define_insn_and_split.
18970 Call ix86_split_simple_return_pop_internal to split it for
18971 -mfunction-return=.
18972 (simple_return_indirect_internal): Call
18973 ix86_output_indirect_function_return instead of
18974 ix86_output_indirect_jmp.
18976 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18979 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18980 memset and value initialization afterwards.
18982 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18984 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18986 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18989 * common/config/aarch64/aarch64-common.c
18990 (aarch_option_optimization_table[]): Switch
18991 off fomit-frame-pointer
18993 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18994 Chung-Ju Wu <jasonwucj@gmail.com>
18996 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18998 (store_multiple): Ditto.
19000 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19002 * config.gcc: Add --with-cpu support for nds32 target.
19003 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
19004 * config/nds32/nds32.opt: Add -mcpu= option.
19006 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
19008 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
19009 isel=yes): Warn for these deprecated options.
19011 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
19013 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
19014 ISA_2_5_MASKS_EMBEDDED.
19016 2018-02-23 Jakub Jelinek <jakub@redhat.com>
19018 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
19019 p->max as pointers rather than using iterative_hash_expr.
19021 2018-02-23 Carl Love <cel@us.ibm.com>
19023 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
19024 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
19026 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
19027 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
19028 P8V_BUILTIN_VEC_VUNSIGNED2.
19030 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
19033 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19034 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
19036 * lra-constraints.c (curr_insn_transform): Set up
19037 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
19039 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
19040 * lra-eliminations.c (spill_pseudos): Ditto.
19041 (process_insn_for_elimination): Ditto.
19042 * lra-lives.c (reg_early_clobber_p): Use the new macros.
19043 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
19044 LRA_NON_CLOBBERED_ALT.
19046 2018-02-22 Martin Sebor <msebor@redhat.com>
19048 PR tree-optimization/84480
19049 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
19050 to maybe_diag_stxncpy_trunc. Call it.
19051 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
19052 from gimple_fold_builtin_strcpy. Print inlining stack.
19053 (handle_builtin_stxncpy): Print inlining stack.
19054 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
19056 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
19059 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
19060 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
19061 and -fcheck-pointer-bounds are used together.
19062 (indirect_thunk_prefix): New enum.
19063 (indirect_thunk_need_prefix): New function.
19064 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
19065 "_nt" instead of "_bnd" for NOTRACK prefix.
19066 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
19067 (output_indirect_thunk_function): Likewise.
19069 (ix86_code_end): Update output_indirect_thunk_function calls.
19070 (ix86_output_indirect_branch_via_reg): Replace
19071 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
19072 (ix86_output_indirect_branch_via_push): Likewise.
19073 (ix86_output_function_return): Likewise.
19074 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
19075 incompatible with -fcf-protection=branch and
19076 -fcheck-pointer-bounds.
19078 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19081 * config/aarch64/aarch64.c (aarch64_print_address_internal):
19082 Change gcc_assert call to output_operand_lossage.
19084 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19086 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
19088 2018-02-22 DJ Delorie <dj@redhat.com>
19089 Sebastian Perta <sebastian.perta@renesas.com>
19090 Oleg Endo <olegendo@gcc.gnu.org>
19092 * config/rx/rx.c (rx_rtx_costs): New function.
19093 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
19095 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
19097 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
19099 2018-02-22 Martin Liska <mliska@suse.cz>
19102 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
19103 Add "native" as a possible value.
19105 2018-02-22 Martin Liska <mliska@suse.cz>
19108 * config/i386/i386.c (ix86_option_override_internal):
19109 Add "native" as a possible value for -march and -mtune.
19111 2018-02-22 Jakub Jelinek <jakub@redhat.com>
19114 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
19115 to all type variants.
19117 PR tree-optimization/84503
19118 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
19119 width as info->bitpos + info->bitsize - start.
19120 (merged_store_group::merge_overlapping): Simplify width computation.
19121 (check_no_overlap): New function.
19122 (imm_store_chain_info::try_coalesce_bswap): Compute expected
19123 start + width and last_order of the group, fail if check_no_overlap
19125 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
19126 to group if check_no_overlap fails.
19128 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19130 * config/rs6000/altivec.md: Delete contraint arguments to
19131 define_expand, define_split, and define_peephole2, and in
19132 define_insn_and_split if always unused.
19133 * config/rs6000/darwin.md: Ditto.
19134 * config/rs6000/dfp.md: Ditto.
19135 * config/rs6000/rs6000.md: Ditto.
19136 * config/rs6000/sync.md: Ditto.
19137 * config/rs6000/vector.md: Ditto.
19138 * config/rs6000/vsx.md: Ditto.
19140 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19142 * config/rs6000/altivec.md: Write output control strings as braced
19143 blocks instead of double-quoted strings.
19144 * config/rs6000/darwin.md: Ditto.
19145 * config/rs6000/rs6000.md: Ditto.
19146 * config/rs6000/vector.md: Ditto.
19147 * config/rs6000/vsx.md: Ditto.
19149 2018-02-21 Jason Merrill <jason@redhat.com>
19151 PR c++/84314 - ICE with templates and fastcall attribute.
19152 * attribs.c (build_type_attribute_qual_variant): Remove assert.
19154 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19156 * ipa-cp.c (determine_versionability): Fix comment typos.
19158 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19161 * ipa-cp.c (determine_versionability): Do not version functions caling
19164 2018-02-21 Martin Liska <mliska@suse.cz>
19167 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
19168 Add "native" as a possible value.
19169 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
19170 the macro when native cpu detection is available.
19172 2018-02-21 Martin Liska <mliska@suse.cz>
19175 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
19176 Add "native" as a possible value.
19177 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
19178 when native cpu detection is available.
19180 2018-02-21 Jakub Jelinek <jakub@redhat.com>
19181 Martin Sebor <msebor@redhat.com>
19183 PR tree-optimization/84478
19184 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
19186 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
19187 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
19188 support which is conservatively correct, for 2 only stay conservative
19189 for maxlen. Formatting and comment capitalization fixes. Add STRICT
19190 argument to the 2 argument get_range_strlen, adjust 6 arg
19191 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
19193 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
19194 (gimple_fold_builtin_strlen): Pass true as last argument to
19197 2018-02-20 Martin Sebor <msebor@redhat.com>
19199 PR middle-end/84095
19200 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
19201 (builtin_memref::set_base_and_offset): Same. Handle inner references.
19202 (builtin_memref::builtin_memref): Factor out parts into
19203 set_base_and_offset and call it.
19205 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19207 PR middle-end/84406
19208 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
19209 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
19210 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
19211 search at the associated MODE_INT.
19213 2018-02-20 Jeff Law <law@redhat.com>
19215 PR middle-end/82123
19216 PR tree-optimization/81592
19217 PR middle-end/79257
19218 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
19219 for range data rather than using global data.
19220 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
19221 range data rather than using global data.
19222 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
19223 pass it to children as needed.
19224 (struct directive::fmtresult): Similarly.
19225 (struct directive::set_width): Similarly.
19226 (struct directive::set_precision): Similarly.
19227 (format_integer, format_directive, parse_directive): Similarly.
19228 (format_none): Accept unnamed vr_values parameter.
19229 (format_percent, format_floating, format_character): Similarly.
19230 (format_string, format_plain): Similarly.
19231 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
19232 the EVRP range analyzer for range data rather than using global data.
19233 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
19234 gimple-ssa-evrp-analyze.h
19235 (class sprintf_dom_walker): Add after_dom_children member function.
19236 Add evrp_range_analyzer member.
19237 (sprintf_dom_walker::before_dom_children): Call into the EVRP
19238 range analyzer as needed.
19239 (sprintf_dom_walker::after_dom_children): New member function.
19240 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
19242 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
19243 (evrp_range_analyzer::pop_to_marker): Likewise.
19245 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19247 PR tree-optimization/84419
19248 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
19249 with the required type if its current type is compatible but
19252 2018-02-20 Jakub Jelinek <jakub@redhat.com>
19254 PR middle-end/82004
19255 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
19256 after vectorization.
19258 2018-02-20 Martin Liska <mliska@suse.cz>
19261 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
19262 possible values if we don't have a hint.
19264 2018-02-20 Martin Liska <mliska@suse.cz>
19268 * final.c (shorten_branches): Build align_tab array with one
19270 * opts.c (finish_options): Add alignment option limit check.
19271 (MAX_CODE_ALIGN): Likewise.
19272 (MAX_CODE_ALIGN_VALUE): Likewise.
19273 * doc/invoke.texi: Document maximum allowed option value for
19274 all -falign-* options.
19276 2018-02-19 Jakub Jelinek <jakub@redhat.com>
19279 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
19280 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
19281 * var-tracking.c (emit_note_insn_var_location): Remove all references
19282 to NOTE_INSN_CALL_ARG_LOCATION.
19283 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
19284 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
19285 Use copy_rtx_if_shared.
19286 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
19287 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
19288 (dwarf2out_var_location): Remove handling of
19289 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
19291 * final.c (final_scan_insn): Remove all references to
19292 NOTE_INSN_CALL_ARG_LOCATION.
19293 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
19294 before dumping final insns.
19295 * except.c (emit_note_eh_region_end): Remove all references to
19296 NOTE_INSN_CALL_ARG_LOCATION.
19297 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
19298 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
19299 * config/arc/arc.c (hwloop_optimize): Likewise.
19300 * config/arm/arm.c (create_fix_barrier): Likewise.
19301 * config/s390/s390.c (s390_chunkify_start): Likewise.
19302 * config/sh/sh.c (find_barrier): Likewise.
19303 * config/i386/i386.c (rest_of_insert_endbranch,
19304 ix86_seh_fixup_eh_fallthru): Likewise.
19305 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
19306 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
19307 * config/frv/frv.c (frv_function_prologue): Likewise.
19308 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
19310 (note_outside_basic_block_p): Remove all references to
19311 NOTE_INSN_CALL_ARG_LOCATION.
19312 * gengtype.c (adjust_field_rtx_def): Likewise.
19313 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
19315 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
19316 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
19319 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
19322 PR tree-optimization/84452
19323 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
19324 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
19327 2018-02-19 Martin Liska <mliska@suse.cz>
19330 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
19332 2018-02-19 Martin Liska <mliska@suse.cz>
19333 Richard Sandiford <richard.sandiford@linaro.org>
19335 PR tree-optimization/82491
19336 * gimple-fold.c (get_base_constructor): Make earlier bail out
19339 2018-02-19 Carl Love <cel@us.ibm.com>
19341 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
19342 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
19344 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
19345 P8V_BUILTIN_VEC_NEG.
19347 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
19349 * config/rl78/rl78.md (movdf): New define expand.
19351 2018-02-19 Martin Liska <mliska@suse.cz>
19354 * doc/invoke.texi: Fix typo.
19355 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
19357 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
19359 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
19360 handle rs6000_single_float and rs6000_double_float specially for
19363 2018-02-16 Jeff Law <law@redhat.com>
19365 * config/rx/rx.c (add_pop_cfi_notes): New function.;
19366 (pop_regs): Use it.
19368 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19371 * ipa-inline.c (inline_small_functions): Fix a typo.
19373 2018-02-16 Nathan Sidwell <nathan@acm.org>
19375 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
19377 2018-02-16 Carl Love <cel@us.ibm.com>
19379 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
19380 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
19381 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
19382 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
19383 expansion to P8V_BUILTIN_VEC_FLOAT2.
19385 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
19387 PR rtl-optimization/70023
19388 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
19389 src_regno into account.
19391 2018-02-16 Carl Love <cel@us.ibm.com>
19393 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
19394 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
19395 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
19396 * config/rs6000/rs6000.c: Remove case statements for
19397 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
19398 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
19399 and P9V_BUILTIN_VEC_VINSERT4B.
19400 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
19401 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
19402 * config/rs6000/vsx.md:
19403 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
19406 2018-02-16 Carl Love <cel@us.ibm.com>
19408 * config/rs6000/altivec.h: Add builtin names vec_extract4b
19410 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
19412 * config/rs6000/rs6000-c.c: Add the definitions for
19413 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
19414 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
19415 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
19416 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
19417 definition for insert4b and define insn *insert3b_internal.
19418 * doc/extend.texi: Add documentation for vec_extract4b.
19420 2018-02-16 Nathan Sidwell <nathan@acm.org>
19422 * doc/extend.texi (Backwards Compatibility): Mention friend
19423 injection. Note for-scope is deprecated.
19424 * doc/invoke.texi (-ffriend-injection): Deprecate.
19426 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
19428 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
19429 that moved to I2, also allow destinations that are a paradoxical
19430 subreg (instead of a normal reg).
19432 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
19435 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
19438 2018-02-16 Richard Biener <rguenther@suse.de>
19440 PR tree-optimization/84037
19441 PR tree-optimization/84016
19443 * config/i386/i386.c (ix86_builtin_vectorization_cost):
19444 Adjust vec_construct for the fact we need additional higher latency
19445 128bit inserts for AVX256 and AVX512 vector builds.
19446 (ix86_add_stmt_cost): Scale vector construction cost for
19449 2018-02-16 Richard Biener <rguenther@suse.de>
19451 PR tree-optimization/84417
19452 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19453 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19454 (non_rewritable_lvalue_p): Likewise, use poly-ints.
19456 2018-02-16 Martin Liska <mliska@suse.cz>
19459 * internal-fn.def (ASAN_CHECK): Set proper flags.
19460 (ASAN_MARK): Likewise.
19462 2018-02-16 Julia Koval <julia.koval@intel.com>
19464 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19465 from PTA_CANNONLAKE.
19467 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19470 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19471 Use ++iter rather than iter++ for std::list iterators.
19472 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
19473 defer deleting them until all nodes in the forest are processed. Do
19474 free even leaf nodes. Change to_process into auto_vec.
19477 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19478 * vec.h (vec_default_construct): Use memset instead of placement new
19479 if BROKEN_VALUE_INITIALIZATION is defined.
19480 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19481 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19484 PR rtl-optimization/83723
19485 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19486 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19487 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19489 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19491 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19493 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19495 PR rtl-optimization/81443
19496 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19497 from inner REGs to paradoxical SUBREGs.
19499 2018-02-16 Richard Biener <rguenther@suse.de>
19501 PR tree-optimization/84399
19502 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19503 For operands we can analyze at their definition make sure we can
19504 analyze them at each use as well.
19506 2018-02-16 Richard Biener <rguenther@suse.de>
19508 PR tree-optimization/84190
19509 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19510 volatile accesses if the decl isn't volatile.
19512 2018-02-15 Jason Merrill <jason@redhat.com>
19514 PR c++/84314 - ICE with templates and fastcall attribute.
19515 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19516 TYPE_CANONICAL on an existing type.
19518 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19520 PR tree-optimization/84383
19521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19522 dstoff nor call operand_equal_p if dstbase is NULL.
19524 PR tree-optimization/84334
19525 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19526 also a CONSTANT_CLASS_P, punt.
19528 2018-02-14 Jim Wilson <jimw@sifive.com>
19530 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19531 first SMALL_OPERAND check. New local min_second_step. Move assert
19532 to where locals are set. Add TARGET_RVC support.
19533 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19535 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19537 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19539 2018-02-14 Martin Sebor <msebor@redhat.com>
19541 PR tree-optimization/83698
19542 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19543 arrays constrain the offset range to their bounds.
19544 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19545 (builtin_access::overlap): Avoid setting the size of overlap if it's
19547 (maybe_diag_overlap): Also consider arrays when deciding what values
19548 of offsets to include in diagnostics.
19550 2018-02-14 Martin Sebor <msebor@redhat.com>
19553 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19554 that correspond to the kind of a declaration.
19556 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19559 * config/pa/pa.md: Load address of PIC label using the linkage table
19560 if the label is nonlocal.
19562 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19564 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19565 warning message if user requests -maltivec=be.
19566 * doc/invoke.texi: Document deprecation of -maltivec=be.
19568 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19571 * config/rs6000/rs6000-c.c: Update definitions for
19572 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19573 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19575 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19578 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19579 add _get_ssp intrinsics. Remove argument from
19580 __builtin_ia32_rdssp[d|q].
19581 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19582 * config/i386/i386-builtin.def: Remove argument from
19583 __builtin_ia32_rdssp[d|q].
19584 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19585 ix86_expand_special_args_builtin for _rdssp[d|q].
19586 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19587 Clear register before usage.
19588 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19589 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19591 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19593 PR tree-optimization/84357
19594 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19595 operand 1 of an ARRAY_REF too.
19597 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19600 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19601 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19603 (set_of_reg): New struct.
19604 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19605 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19606 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19608 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19609 Split into bitclr, bitset, bitinvert patterns if appropriate.
19610 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19611 use rx_fuse_in_memory_bitop.
19612 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19613 to named insn, correct maximum insn length.
19615 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19618 * machmode.def: Define a complex mode for PARTIAL_INT.
19619 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19621 * doc/rtl.texi: Document CSPImode.
19622 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19624 (msp430_hard_regno_nregs_with_padding): Likewise.
19626 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19629 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19631 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19633 PR rtl-optimization/84169
19634 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19635 we generated a parallel as new i3 and we split that to new i2 and i3
19636 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19637 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19638 those to i2, not i1. Partially rewrite this scan code.
19640 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19643 * stor-layout.c (place_field): For variable length fields, adjust
19644 offset_align afterwards not just based on the field's alignment,
19645 but also on the size.
19647 PR middle-end/84309
19648 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19649 of exps and logs in the use_exp2 case.
19651 2018-02-13 Jeff Law <law@redhat.com>
19653 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19654 entry for "vector".
19656 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19659 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19663 * common.opt (gas-loc-support, gas-locview-support): New.
19664 (ginline-points, ginternal-reset-location-views): New.
19665 * doc/invoke.texi: Document them. Use @itemx where intended.
19666 (gvariable-location-views): Adjust.
19667 * target.def (reset_location_view): New.
19668 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19669 (TARGET_RESET_LOCATION_VIEW): New.
19670 * doc/tm.texi: Rebuilt.
19671 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19672 (dwarf2out_default_as_locview_support): New.
19673 (output_asm_line_debug_info): Use option variables.
19674 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19675 (output_loc_list): Likewise.
19676 (add_high_low_attributes): Check option variables.
19677 Don't output entry view attribute in strict mode.
19678 (gen_inlined_subroutine_die): Check option variables.
19679 (dwarf2out_inline_entry): Likewise.
19680 (init_sections_and_labels): Likewise.
19681 (dwarf2out_early_finish): Likewise.
19682 (maybe_reset_location_view): New, from...
19683 (dwarf2out_var_location): ... here. Call it.
19684 * debug.h (dwarf2out_default_as_loc_support): Declare.
19685 (dwarf2out_default_as_locview_support): Declare.
19686 * hooks.c (hook_int_rtx_insn_0): New.
19687 * hooks.h (hook_int_rtx_insn_0): Declare.
19688 * toplev.c (process_options): Take -gas-loc-support and
19689 -gas-locview-support from dwarf2out. Enable
19690 -gvariable-location-views by default only with locview
19691 assembler support. Enable -ginternal-reset-location-views by
19692 default only if the target defines the corresponding hook.
19693 Enable -ginline-points by default if location views are
19694 enabled; force it disabled if statement frontiers are
19696 * tree-inline.c (expand_call_inline): Check option variables.
19697 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19699 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
19701 PR tree-optimization/84321
19702 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19703 handling. Also check whether the anti-range contains any values
19704 that satisfy the mask; switch to a VR_RANGE if not.
19706 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
19709 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19711 2018-02-13 Martin Jambor <mjambor@suse.cz>
19714 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19715 of call statements, also set location of a load to a temporary.
19717 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
19719 * config/rl78/rl78.c (add_vector_labels): New function.
19720 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19721 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19722 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19723 which checks that no arguments are passed.
19724 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19725 * doc/extend.texi: Documentation for the new attribute.
19727 2018-02-13 Andreas Schwab <schwab@suse.de>
19729 * config/riscv/linux.h (CPP_SPEC): Define.
19731 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19734 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19735 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19736 OPTION_MASK_ISA_AES as first argument to def_builtin_const
19737 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19738 instead of OPTION_MASK_ISA_PCLMUL as first argument to
19739 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19740 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19741 temporarily for AES and PCLMUL builtins.
19743 PR tree-optimization/84339
19744 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19745 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19748 PR middle-end/84309
19749 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19750 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19751 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19753 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19755 * omp-simd-clone.h: New file.
19756 * omp-simd-clone.c: Include omp-simd-clone.h.
19757 (expand_simd_clones): No longer static.
19758 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19759 cgraph.h and omp-simd-clone.h.
19760 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19761 (vect_recog_widen_shift_pattern): Formatting fix.
19762 (vect_pattern_recog_1): Don't check optab for calls.
19765 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19766 operands[2] into a REG before using gen_lowpart on it.
19768 2018-02-12 Jeff Law <law@redhat.com>
19771 * config/sh/sh.c (find_barrier): Consider a sibling call
19774 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19775 successfully back substituting a reg.
19777 2018-02-12 Richard Biener <rguenther@suse.de>
19779 PR tree-optimization/84037
19780 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19781 parameter, move visited init to caller.
19782 (vect_slp_analyze_operations): Separate cost from validity
19783 check, initialize visited once for all instances.
19784 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19786 * tree-vect-stmts.c (vect_model_simple_cost): Make early
19788 (vect_model_promotion_demotion_cost): Likewise.
19789 (vectorizable_bswap): Guard cost modeling with !slp_node
19790 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19792 (vectorizable_call): Likewise.
19793 (vectorizable_conversion): Likewise.
19794 (vectorizable_assignment): Likewise.
19795 (vectorizable_shift): Likewise.
19796 (vectorizable_operation): Likewise.
19797 (vectorizable_store): Likewise.
19798 (vectorizable_load): Likewise.
19799 (vectorizable_condition): Likewise.
19800 (vectorizable_comparison): Likewise.
19802 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
19805 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19806 (ASAN_MARK): Fix fnspec to account for return value, change pointer
19807 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19809 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19811 PR middle-end/83665
19812 * params.def (inline-min-speedup): Increase from 8 to 15.
19813 (max-inline-insns-auto): Decrease from 40 to 30.
19814 * ipa-split.c (consider_split): Add some buffer for function to
19815 be considered inlining candidate.
19816 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19819 2018-02-12 Richard Biener <rguenther@suse.de>
19821 PR tree-optimization/84037
19822 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19823 matched stmts if we cannot swap the non-matched ones.
19825 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19827 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19828 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19829 _mm_maskz_scalef_round_ss): New intrinsics.
19830 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19831 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19832 __builtin_ia32_scalefss_round): Remove.
19833 (__builtin_ia32_scalefsd_mask_round,
19834 __builtin_ia32_scalefss_mask_round): New intrinsics.
19835 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19836 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19837 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19838 "<round_constraint>")): Changed to ...
19839 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19840 "<round_scalar_constraint>")): ... this.
19841 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19842 %0, %1, %2<round_op3>}"): Changed to ...
19843 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19844 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19845 %2<round_scalar_mask_op3>}"): ... this.
19846 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19848 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19850 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19851 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19852 (_mm_maskz_sqrt_round_ss): New intrinsics.
19853 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19854 (__builtin_ia32_sqrtsd_mask_round)
19855 (__builtin_ia32_sqrtss_mask_round): New builtins.
19856 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19857 (__builtin_ia32_sqrtss_round): Remove.
19858 (__builtin_ia32_sqrtsd_mask_round)
19859 (__builtin_ia32_sqrtss_mask_round): New builtins.
19860 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19861 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19862 ((match_operand:VF_128 1 "vector_operand"
19863 "xBm,<round_constraint>")): Changed to ...
19864 ((match_operand:VF_128 1 "vector_operand"
19865 "xBm,<round_scalar_constraint>")): ... this.
19866 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19867 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19868 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19869 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19870 %<iptr>1<round_scalar_mask_op3>}): ... this.
19871 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19872 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19874 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19877 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19878 Cast vec_cmpeq result to correct type.
19879 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19880 Cast vec_cmpgt result to correct type.
19882 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19884 * final.c (final_scan_insn_1): Renamed from...
19885 (final_scan_insn): ... this. New wrapper, to recover
19886 seen from the outermost call in recursive ones.
19887 * config/sparc/sparc.c (output_return): Drop seen from call.
19888 (output_sibcall): Likewise.
19889 * config/visium/visium.c (output_branch): Likewise.
19891 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19893 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19896 2018-02-10 Alan Modra <amodra@gmail.com>
19899 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19900 Specify LR as an input.
19902 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19905 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19906 remove_member_access_dummy_vars): New functions.
19907 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19908 lower_omp_1, execute_lower_omp): Use them.
19910 PR rtl-optimization/84308
19911 * shrink-wrap.c (spread_components): Release todo vector.
19913 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19915 PR rtl-optimization/57193
19916 * ira-color.c (struct allocno_color_data): Add member
19917 conflict_allocno_hard_prefs.
19918 (update_conflict_allocno_hard_prefs): New.
19919 (bucket_allocno_compare_func): Add a preference based on
19920 conflict_allocno_hard_prefs.
19921 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19922 (color_allocnos): Remove a dead code. Initiate
19923 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19925 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19928 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19929 constraint from =wa to wa. Avoid a subreg on the output operand,
19930 instead use a pseudo and subreg it in a move.
19931 (p9_xxbrd_<mode>): Changed to ...
19932 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19933 (p9_xxbrd_v2df): New expander.
19934 (p9_xxbrw_<mode>): Changed to ...
19935 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19936 (p9_xxbrw_v4sf): New expander.
19938 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19940 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19942 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19945 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19946 multiply in 32-bit mode.
19947 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19948 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19951 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19953 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19954 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19955 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19956 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19958 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19961 * dwarf2out.c (is_trivial_indirect_ref): New function.
19962 (dwarf2out_late_global_decl): Do not generate a location
19963 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19964 and that are not defined in the current unit.
19966 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19968 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19969 instead of a libcall for UNORDERED.
19971 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19974 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19975 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19977 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19980 * config/s390/s390.c (s390_set_current_function): Invoke
19981 s390_indirect_branch_settings also if fndecl didn't change.
19983 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19985 * config/rs6000/rs6000.md (blockage): Set length to zero.
19987 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19989 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19991 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19994 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19995 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19999 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
20000 PARALLEL incoming that failed vt_get_decl_and_offset check.
20002 PR middle-end/84237
20003 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
20004 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
20006 (get_variable_section): For decls in named .bss* sections pass true as
20007 second argument to bss_initializer_p.
20009 2018-02-09 Marek Polacek <polacek@redhat.com>
20010 Jakub Jelinek <jakub@redhat.com>
20013 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
20014 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
20015 Sync some changes from cxx_fold_indirect_ref.
20017 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20019 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
20021 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
20022 (BLOCK_INLINE_ENTRY_LABEL): New.
20023 (dwarf2out_var_location): Disregard inline entry markers.
20024 (inline_entry_data): New struct.
20025 (inline_entry_data_hasher): New hashtable type.
20026 (inline_entry_data_hasher::hash): New.
20027 (inline_entry_data_hasher::equal): New.
20028 (inline_entry_data_table): New variable.
20029 (add_high_low_attributes): Add DW_AT_entry_pc and
20030 DW_AT_GNU_entry_view attributes if a pending entry is found
20031 in inline_entry_data_table. Add old entry_pc attribute only
20032 if debug nonbinding markers are disabled.
20033 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
20034 markers are enabled.
20035 (block_within_block_p, dwarf2out_inline_entry): New.
20036 (dwarf2out_finish): Check that no entries remained in
20037 inline_entry_data_table.
20038 * final.c (reemit_insn_block_notes): Handle inline entry notes.
20039 (final_scan_insn, notice_source_line): Likewise.
20040 (rest_of_clean_state): Skip inline entry markers.
20041 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
20043 * gimple.c (gimple_build_debug_inline_entry): New.
20044 * gimple.h (enum gimple_debug_subcode): Add
20045 GIMPLE_DEBUG_INLINE_ENTRY.
20046 (gimple_build_debug_inline_entry): Declare.
20047 (gimple_debug_inline_entry_p): New.
20048 (gimple_debug_nonbind_marker_p): Adjust.
20049 * insn-notes.def (INLINE_ENTRY): New.
20050 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
20051 inline entry marker notes.
20052 (print_insn): Likewise.
20053 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
20054 (INSN_DEBUG_MARKER_KIND): Likewise.
20055 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
20056 * tree-inline.c (expand_call_inline): Build and insert
20057 debug_inline_entry stmt.
20058 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
20059 inline entry blocks early, if nonbind markers are enabled.
20060 (dump_scope_block): Dump fragment info.
20061 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
20062 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
20063 (gimple_build_debug_inline_entry): New.
20064 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
20065 Enable/disable inline entry points too.
20066 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
20067 (DEBUG_INSN): Describe inline entry markers.
20069 * common.opt (gvariable-location-views): New.
20070 (gvariable-location-views=incompat5): New.
20071 * config.in: Rebuilt.
20072 * configure: Rebuilt.
20073 * configure.ac: Test assembler for view support.
20074 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
20075 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
20076 * dwarf2out.c (var_loc_view): New typedef.
20077 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
20078 (dwarf2out_locviews_in_attribute): New.
20079 (dwarf2out_locviews_in_loclist): New.
20080 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
20081 (enum dw_line_info_opcode): Add LI_adv_address.
20082 (struct dw_line_info_table): Add view.
20083 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
20084 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
20085 (zero_view_p): New variable.
20086 (ZERO_VIEW_P): New macro.
20087 (output_asm_line_debug_info): New.
20088 (struct var_loc_node): Add view.
20089 (add_AT_view_list, AT_loc_list): New.
20090 (add_var_loc_to_decl): Add view param. Test it against last.
20091 (new_loc_list): Add view params. Record them.
20092 (AT_loc_list_ptr): Handle loc and view lists.
20093 (view_list_to_loc_list_val_node): New.
20094 (print_dw_val): Handle dw_val_class_view_list.
20095 (size_of_die): Likewise.
20096 (value_format): Likewise.
20097 (loc_list_has_views): New.
20098 (gen_llsym): Set vl_symbol too.
20099 (maybe_gen_llsym, skip_loc_list_entry): New.
20100 (dwarf2out_maybe_output_loclist_view_pair): New.
20101 (output_loc_list): Output view list or entries too.
20102 (output_view_list_offset): New.
20103 (output_die): Handle dw_val_class_view_list.
20104 (output_dwarf_version): New.
20105 (output_compilation_unit_header): Use it.
20106 (output_skeleton_debug_sections): Likewise.
20107 (output_rnglists, output_line_info): Likewise.
20108 (output_pubnames, output_aranges): Update version comments.
20109 (output_one_line_info_table): Output view numbers in asm comments.
20110 (dw_loc_list): Determine current endview, pass it to new_loc_list.
20111 Call maybe_gen_llsym.
20112 (loc_list_from_tree_1): Adjust.
20113 (add_AT_location_description): Create view list attribute if
20114 needed, check it's absent otherwise.
20115 (convert_cfa_to_fb_loc_list): Adjust.
20116 (maybe_emit_file): Call output_asm_line_debug_info for test.
20117 (dwarf2out_var_location): Reset views as needed. Precompute
20118 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
20119 attribute. Set view.
20120 (new_line_info_table): Reset next view.
20121 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
20122 (dwarf2out_source_line): Likewise. Output view resets and labels to
20123 the assembler, or select appropriate line info opcodes.
20124 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
20125 (optimize_string_length): Catch it. Adjust.
20126 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
20127 dw_val_class_view_list, and remove it if no longer needed.
20128 (hash_loc_list): Hash view numbers.
20129 (loc_list_hasher::equal): Compare them.
20130 (optimize_location_lists): Check whether a view list symbol is
20131 needed, and whether the locview attribute is present, and
20132 whether they match. Remove the locview attribute if no longer
20134 (index_location_lists): Call skip_loc_list_entry for test.
20135 (dwarf2out_finish): Call output_asm_line_debug_info for test.
20136 Use output_dwarf_version.
20137 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
20138 (struct dw_val_node): Add val_view_list.
20139 * final.c (SEEN_NEXT_VIEW): New.
20140 (set_next_view_needed): New.
20141 (clear_next_view_needed): New.
20142 (maybe_output_next_view): New.
20143 (final_start_function): Rename to...
20144 (final_start_function_1): ... this. Take pointer to FIRST,
20145 add SEEN parameter. Emit param bindings in the initial view.
20146 (final_start_function): Reintroduce SEEN-less interface.
20147 (final): Rename to...
20148 (final_1): ... this. Take SEEN parameter. Output final pending
20149 next view at the end.
20150 (final): Reintroduce seen-less interface.
20151 (final_scan_insn): Output pending next view before switching
20152 sections or ending a block. Mark the next view as needed when
20153 outputting variable locations. Notify debug backend of section
20154 changes, and of location view changes.
20155 (rest_of_handle_final): Adjust.
20156 * toplev.c (process_options): Autodetect value for debug variable
20157 location views option. Warn on incompat5 without -gdwarf-5.
20158 * doc/invoke.texi (gvariable-location-views): New.
20159 (gvariable-location-views=incompat5): New.
20160 (gno-variable-location-views): New.
20162 2018-02-08 David Malcolm <dmalcolm@redhat.com>
20164 PR tree-optimization/84136
20165 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
20166 that the result of find_edge is non-NULL.
20168 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20171 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
20172 storing integer register in SImode. Fix cost of 256 and 512
20173 byte aligned SSE register store.
20175 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20177 * config/i386/i386.c (ix86_multiplication_cost): Fix
20178 multiplication cost for TARGET_AVX512DQ.
20180 2018-02-08 Marek Polacek <polacek@redhat.com>
20182 PR tree-optimization/84238
20183 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
20186 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20188 PR tree-optimization/84265
20189 * tree-vect-stmts.c (vectorizable_store): Don't treat
20190 VMAT_CONTIGUOUS accesses as grouped.
20191 (vectorizable_load): Likewise.
20193 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20195 PR tree-optimization/81635
20196 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
20197 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
20198 (test_round_for_mask): New functions.
20199 (wide_int_cc_tests): Call test_round_for_mask.
20200 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
20201 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
20202 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
20203 range returned by get_range_info.
20205 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20208 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
20209 * symtab.c: Include builtins.h
20210 (symtab_node::output_to_lto_symbol_table_p): Move here
20211 from lto-streamer-out.c:output_symbol_p.
20212 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
20213 (output_symbol_p): Move all logic to symtab.c
20214 (produce_symtab): Update.
20216 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20218 * config/s390/s390-opts.h (enum indirect_branch): Define.
20219 * config/s390/s390-protos.h (s390_return_addr_from_memory)
20220 (s390_indirect_branch_via_thunk)
20221 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
20222 (enum s390_indirect_branch_type): Define.
20223 * config/s390/s390.c (struct s390_frame_layout, struct
20224 machine_function): Remove.
20225 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
20226 (indirect_branch_table_label_no, indirect_branch_table_name):
20228 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
20229 (enum s390_indirect_branch_option): Define.
20230 (s390_return_addr_from_memory): New function.
20231 (s390_handle_string_attribute): New function.
20232 (s390_attribute_table): Add new attribute handler.
20233 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
20234 (s390_indirect_branch_via_thunk): New function.
20235 (s390_indirect_branch_via_inline_thunk): New function.
20236 (s390_function_ok_for_sibcall): When jumping via thunk disallow
20237 sibling call optimization for non z10 compiles.
20238 (s390_emit_call): Force indirect branch target to be a single
20239 register. Add r1 clobber for non-z10 compiles.
20240 (s390_emit_epilogue): Emit return jump via return_use expander.
20241 (s390_reorg): Handle JUMP_INSNs as execute targets.
20242 (s390_option_override_internal): Perform validity checks for the
20243 new command line options.
20244 (s390_indirect_branch_attrvalue): New function.
20245 (s390_indirect_branch_settings): New function.
20246 (s390_set_current_function): Invoke s390_indirect_branch_settings.
20247 (s390_output_indirect_thunk_function): New function.
20248 (s390_code_end): Implement target hook.
20249 (s390_case_values_threshold): Implement target hook.
20250 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
20252 * config/s390/s390.h (struct s390_frame_layout)
20253 (struct machine_function): Move here from s390.c.
20254 (TARGET_INDIRECT_BRANCH_NOBP_RET)
20255 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
20256 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
20257 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
20258 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
20259 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
20260 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
20261 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
20262 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
20263 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
20264 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
20265 (mnemonic attribute): Add values which aren't recognized
20267 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
20268 pattern for branch conversion. Fix mnemonic attribute.
20269 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
20270 indirect branch via thunk if requested.
20271 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
20272 ("*indirect_jump"): Disable for branch conversion using out of
20274 ("indirect_jump_via_thunk<mode>_z10")
20275 ("indirect_jump_via_thunk<mode>")
20276 ("indirect_jump_via_inlinethunk<mode>_z10")
20277 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
20278 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
20279 ("casesi_jump_via_inlinethunk<mode>_z10")
20280 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
20281 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
20282 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
20283 ("*indirect2_jump"): Disable for branch conversion.
20284 ("casesi_jump"): Turn into expander and expand patterns for branch
20286 ("return_use"): New expander.
20287 ("*return"): Emit return via thunk and rename it to ...
20288 ("*return<mode>"): ... this one.
20289 * config/s390/s390.opt: Add new options and and enum for the
20292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20294 * lra-constraints.c (match_reload): Unconditionally use
20295 gen_lowpart_SUBREG, rather than selecting between that
20296 and equivalent gen_rtx_SUBREG code.
20298 2018-02-08 Richard Biener <rguenther@suse.de>
20300 PR tree-optimization/84233
20301 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
20302 changed flag instead of boguously re-using phi_inserted.
20304 2018-02-08 Martin Jambor <mjambor@suse.cz>
20306 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
20307 static local variables.
20309 2018-02-08 Richard Biener <rguenther@suse.de>
20311 PR tree-optimization/84278
20312 * tree-vect-stmts.c (vectorizable_store): When looking for
20313 smaller vector types to perform grouped strided loads/stores
20314 make sure the mode is supported by the target.
20315 (vectorizable_load): Likewise.
20317 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20319 * config/aarch64/aarch64.c (aarch64_components_for_bb):
20320 Increase LDP/STP opportunities by adding adjacent callee-saves.
20322 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20324 PR rtl-optimization/84068
20325 PR rtl-optimization/83459
20326 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
20328 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
20330 PR tree-optimization/84224
20331 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
20332 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
20333 non-zero arguments.
20335 2018-02-07 Iain Sandoe <iain@codesourcery.com>
20338 * config/rs6000/altivec.md (*restore_world): Remove LR use.
20339 * config/rs6000/predicates.md (restore_world_operation): Adjust op
20340 count, remove one USE.
20342 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
20344 * doc/install.texi (Configuration): Document the
20345 --with-long-double-format={ibm,ieee} PowerPC configuration
20349 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
20350 Convert from define_expand to be define_insn_and_split. Rework
20351 float/double/_Float128 conversions to QI/HI/SImode to work with
20352 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
20353 conversions to QI/HImode types did a store and then a load to
20354 truncate the value. For conversions to VSX registers, don't split
20355 the insn, instead emit the code directly. Use the code iterator
20356 any_fix to combine signed and unsigned conversions.
20357 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
20358 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
20359 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20360 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
20361 (fix_<mode>di2_hw): Likewise.
20362 (fixuns_<mode>di2_hw): Likewise.
20363 (fix_<mode>si2_hw): Likewise.
20364 (fixuns_<mode>si2_hw): Likewise.
20365 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
20366 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20367 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
20368 fix<uns>_trunc<SFDF:mode>si2_p8.
20369 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
20370 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
20371 (fix<uns>_<mode>_mem): Likewise.
20372 (fctiw<u>z_<mode>_mem): Likewise.
20373 (fix<uns>_<mode>_mem): Likewise.
20374 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
20375 the register allocator from doing a direct move to the GPRs to do
20376 a store, and instead use the ISA 3.0 store byte/half-word from
20377 vector register instruction. For IEEE 128-bit floating point,
20378 also optimize stores of 32-bit ints.
20379 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
20381 2018-02-07 Alan Hayward <alan.hayward@arm.com>
20383 * genextract.c (push_pathstr_operand): New function to support
20385 (walk_rtx): Call push_pathstr_operand.
20386 (print_path): Support [a-zA-Z].
20388 2018-02-07 Richard Biener <rguenther@suse.de>
20390 PR tree-optimization/84037
20391 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
20392 (cse_and_gimplify_to_preheader): Declare.
20393 (vect_get_place_in_interleaving_chain): Likewise.
20394 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20396 (_loop_vec_info::~_loop_vec_info): Delete it.
20397 (cse_and_gimplify_to_preheader): New function.
20398 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
20399 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
20400 (vectorizable_load): Likewise. For grouped stores always base
20401 the IV on the first element.
20402 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
20403 condition before gimplifying.
20405 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20407 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
20408 *DIV_EXPR and *MOD_EXPR.
20410 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
20413 * config/i386/i386.c (ix86_option_override_internal): Mask out
20414 the CF_SET bit when checking -fcf-protection.
20416 2018-02-07 Tom de Vries <tom@codesourcery.com>
20419 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
20422 2018-02-07 Richard Biener <rguenther@suse.de>
20424 PR tree-optimization/84204
20425 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
20428 PR tree-optimization/84205
20429 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
20430 special-case isl_ast_op_zdiv_r.
20432 PR tree-optimization/84223
20433 * graphite-scop-detection.c (gather_bbs::before_dom_children):
20434 Only add conditions from within the region.
20435 (gather_bbs::after_dom_children): Adjust.
20437 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
20440 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
20441 * config/avr/avr.md: Only post-reload split REG-REG moves if
20442 either register is GENERAL_REG_P.
20444 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20446 PR tree-optimization/84235
20447 * tree-ssa-scopedtables.c
20448 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20449 if the subtraction is performed in floating point type where NaNs are
20450 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20451 build 1. Formatting fix.
20453 2018-02-06 Jakub Jelinek <jakub@redhat.com>
20456 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20457 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20458 and skip it regardless of bb boundaries. Use CALL_P macro,
20459 don't test INSN_P (insn) together with CALL_P or JUMP_P check
20460 unnecessarily, formatting fix.
20462 2018-02-06 Michael Collison <michael.collison@arm.com>
20464 * config/arm/thumb2.md:
20465 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20466 (*thumb_mov_notscc): Ditto.
20468 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
20471 * config/rs6000/rs6000.md (su code attribute): Use "u" for
20472 unsigned_fix, not "s".
20474 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20476 * configure.ac (gcc_fn_eh_frame_ro): New function.
20477 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20478 correct .eh_frame permissions.
20479 * configure: Regenerate.
20481 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20483 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20484 irrelevant options.
20486 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20488 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20489 Display warning message for -mno-speculate-indirect-jumps.
20491 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20493 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20495 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20497 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20499 PR tree-optimization/84225
20500 * tree-eh.c (find_trapping_overflow): Only call
20501 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20503 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20506 * config/i386/i386.c: Reimplement the check of possible options
20507 -mibt/-mshstk conbination. Change error messages.
20508 * doc/invoke.texi: Fix a typo: remove extra '='.
20510 2018-02-06 Marek Polacek <polacek@redhat.com>
20512 PR tree-optimization/84228
20513 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20515 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20518 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20519 emitted arch directives.
20520 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20521 __ARM_FEATURE_COPROC before changing architectures.
20523 2018-02-06 Richard Biener <rguenther@suse.de>
20525 * config/i386/i386.c (print_reg): Fix typo.
20526 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20528 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20530 * configure: Regenerate.
20532 2018-02-05 Martin Sebor <msebor@redhat.com>
20534 PR tree-optimization/83369
20535 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20538 2018-02-05 Martin Liska <mliska@suse.cz>
20540 * doc/invoke.texi: Cherry-pick upstream r323995.
20542 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20544 * ira.c (ira_init_register_move_cost): Adjust comment.
20546 2018-02-05 Martin Liska <mliska@suse.cz>
20548 PR gcov-profile/84137
20549 * doc/gcov.texi: Fix typo in documentation.
20551 2018-02-05 Martin Liska <mliska@suse.cz>
20553 PR gcov-profile/83879
20554 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20555 using dlopen functionality.
20557 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20559 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20560 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20561 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20562 _mm_maskz_range_round_ss): New intrinsics.
20563 (__builtin_ia32_rangesd128_round)
20564 (__builtin_ia32_rangess128_round): Remove.
20565 (__builtin_ia32_rangesd128_mask_round,
20566 __builtin_ia32_rangess128_mask_round): New builtins.
20567 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20568 __builtin_ia32_rangess128_round): Remove.
20569 (__builtin_ia32_rangesd128_mask_round,
20570 __builtin_ia32_rangess128_mask_round): New builtins.
20571 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20572 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20573 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20574 "<round_saeonly_constraint>")): Changed to ...
20575 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20576 "<round_saeonly_scalar_constraint>")): ... this.
20577 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20578 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20579 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20580 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20581 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20583 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20585 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20587 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20588 Remove all values except native, 8540 and 8548.
20590 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20592 * config/i386/i386.c (ix86_output_function_return): Pass
20593 INVALID_REGNUM, instead of -1, as invalid register number to
20594 indirect_thunk_name and output_indirect_thunk.
20596 2018-02-02 Julia Koval <julia.koval@intel.com>
20598 * config.gcc: Add -march=icelake.
20599 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20600 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20601 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20602 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20603 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20604 (processor_target_table): Add icelake.
20605 (ix86_option_override_internal): Handle new PTAs.
20606 (get_builtin_code_for_version): Handle icelake.
20607 (M_INTEL_COREI7_ICELAKE): New.
20608 (fold_builtin_cpu): Handle icelake.
20609 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20610 * doc/invoke.texi: Add -march=icelake.
20612 2018-02-02 Julia Koval <julia.koval@intel.com>
20614 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20615 to wide_int_bitmask.
20616 * wide-int-bitmask.h: New.
20618 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20621 * config/i386/i386.md: Replace Pmode with word_mode in
20622 builtin_setjmp_setup and builtin_longjmp to support x32.
20624 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20628 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20630 (rs6000_supported_cpu_names): New static variable.
20631 (linux_cpu_translation_table): Likewise.
20632 (elf_platform) <cpu>: Define new static variable and use it.
20633 Translate kernel AT_PLATFORM name to canonical name if needed.
20634 Error if platform name is unknown.
20636 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20639 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20641 2018-02-01 Jeff Law <law@redhat.com>
20644 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20645 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20646 the scratch if RELEASE_VIA_POP is false.
20647 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20648 If we have to save a temporary register, decrement SIZE appropriately.
20649 Pass new arguments to release_scratch_register_on_entry.
20650 (ix86_adjust_stack_and_probe): Likewise.
20651 (ix86_emit_probe_stack_range): Pass new arguments to
20652 release_scratch_register_on_entry.
20654 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20656 PR rtl-optimization/84157
20657 * combine.c (change_zero_ext): Use REG_P predicate in
20658 front of HARD_REGISTER_P predicate.
20660 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20662 * config/avr/avr.c (avr_option_override): Move disabling of
20663 -fdelete-null-pointer-checks to...
20664 * common/config/avr/avr-common.c (avr_option_optimization_table):
20667 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20669 PR tree-optimization/81635
20670 * tree-data-ref.c (split_constant_offset_1): For types that
20671 wrap on overflow, try to use range info to prove that wrapping
20674 2018-02-01 Renlin Li <renlin.li@arm.com>
20677 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20678 TAILCALL_ADDR_REGS.
20679 (aarch64_register_move_cost): Likewise.
20680 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20681 TAILCALL_ADDR_REGS.
20682 (REG_CLASS_NAMES): Likewise.
20683 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20684 TAILCALL_ADDR_REGS. Remove IP registers.
20685 * config/aarch64/aarch64.md (Ucs): Update register constraint.
20687 2018-02-01 Richard Biener <rguenther@suse.de>
20689 * domwalk.h (dom_walker::dom_walker): Add additional constructor
20690 for specifying RPO order and allow NULL for that.
20691 * domwalk.c (dom_walker::dom_walker): Likewise.
20692 (dom_walker::walk): Handle NULL RPO order.
20693 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20695 (rewrite_update_dom_walker): Likewise.
20696 (mark_def_dom_walker): Likewise.
20698 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20700 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20701 (aarch64_maybe_expand_sve_subreg_move): Declare.
20702 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20703 * config/aarch64/predicates.md (aarch64_any_register_operand): New
20705 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20706 that are semantically a reverse operation.
20707 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20708 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20709 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20711 (aarch64_can_change_mode_class): For big-endian, forbid changes
20712 between two SVE modes if they have different element sizes.
20714 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20716 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20717 the TImode handling for big-endian targets.
20719 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20721 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20722 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
20724 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20725 Remove BSWAP handing for big-endian targets and use the form of
20726 LD1RQ appropariate for the mode.
20728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20730 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20731 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20732 duplicated element.
20734 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20737 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20738 check for operands that need to go through aarch64_sve_reload_be.
20740 2018-02-01 Jakub Jelinek <jakub@redhat.com>
20742 PR tree-optimization/81661
20743 PR tree-optimization/84117
20744 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20745 * tree-eh.c: Include gimplify.h.
20746 (find_trapping_overflow, replace_trapping_overflow,
20747 rewrite_to_non_trapping_overflow): New functions.
20748 * tree-vect-loop.c: Include tree-eh.h.
20749 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20750 * tree-data-ref.c: Include tree-eh.h.
20751 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20753 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
20755 PR rtl-optimization/84123
20756 * combine.c (change_zero_ext): Check if hard register satisfies
20757 can_change_dest_mode before calling gen_lowpart_SUBREG.
20759 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
20762 * ira.c (ira_init_register_move_cost): Remove assert.
20764 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20766 PR rtl-optimization/84071
20767 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20768 * doc/tm.texi: Regenerate.
20770 2018-01-31 Richard Biener <rguenther@suse.de>
20772 PR tree-optimization/84132
20773 * tree-data-ref.c (analyze_miv_subscript): Properly
20774 check whether evolution_function_is_affine_multivariate_p
20775 before calling gcd_of_steps_may_divide_p.
20777 2018-01-31 Julia Koval <julia.koval@intel.com>
20780 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20781 * config/i386/i386.md (rdpid_rex64) New.
20782 (rdpid): Make 32bit only.
20784 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
20787 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20788 an IDENTIFIER_NODE for FUNCTION_TYPE's.
20790 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20793 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20795 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20797 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20799 PR rtl-optimization/84071
20800 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20801 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20803 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20805 * config/arc/arc.c (arc_handle_aux_attribute): New function.
20806 (arc_attribute_table): Add 'aux' attribute.
20807 (arc_in_small_data_p): Consider aux like variables.
20808 (arc_is_aux_reg_p): New function.
20809 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20810 (arc_get_aux_arg): New function.
20811 (prepare_move_operands): Handle aux-register access.
20812 (arc_handle_aux_attribute): New function.
20813 * doc/extend.texi (ARC Variable attributes): Add subsection.
20815 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20817 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20818 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20819 (arc_attribute_table): Add 'uncached' attribute.
20820 (arc_print_operand): Print '.di' flag for uncached memory
20822 (arc_in_small_data_p): Do not consider for small data the uncached
20824 (arc_is_uncached_mem_p): New function.
20825 * config/arc/predicates.md (compact_store_memory_operand): Check
20826 for uncached memory accesses.
20827 (nonvol_nonimm_operand): Likewise.
20828 * doc/extend.texi (ARC Type Attribute): New subsection.
20830 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20833 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20834 falign-loops=): Add Optimization flag.
20836 2018-01-30 Jeff Law <law@redhat.com>
20839 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20840 INT_REGISTERS_SAVED. Check it prior to calling
20841 get_scratch_register_on_entry.
20842 (ix86_adjust_stack_and_probe): Similarly.
20843 (ix86_emit_probe_stack_range): Similarly.
20844 (ix86_expand_prologue): Corresponding changes.
20846 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20849 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20850 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20852 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20855 * lra-constraints.c (curr_insn_transform): Process AND in the
20858 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20860 PR rtl-optimization/83986
20861 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20862 dependence against last_pending_memory_flush in addition to
20863 pending_jump_insns.
20865 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20867 PR tree-optimization/81611
20868 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20871 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20874 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20877 2018-01-30 Richard Biener <rguenther@suse.de>
20878 Jakub Jelinek <jakub@redhat.com>
20880 PR tree-optimization/84111
20881 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20882 inner loops added during recursion, as they don't have up-to-date
20885 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20888 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20889 (can_inline_edge_by_limits_p): ... here.
20890 (can_early_inline_edge_p, check_callers,
20891 update_caller_keys, update_callee_keys, recursive_inlining,
20892 add_new_edges_to_heap, speculation_useful_p,
20893 inline_small_functions,
20894 inline_small_functions, flatten_function,
20895 inline_to_all_callers_1): Update.
20897 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20899 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20902 2018-01-30 Richard Biener <rguenther@suse.de>
20904 PR tree-optimization/83008
20905 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20906 invariant and constant vector uses in stmts when they need
20907 more than one stmt.
20909 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20912 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20913 * configure: Regenerate.
20915 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20917 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20919 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20920 Use gen_rtx_REG rather than gen_lowpart.
20922 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20924 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20925 rather than 0 when creating partial subregs.
20927 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20929 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20932 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20935 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20936 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20937 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20938 flags. This restores the settings used before the 2017-07-24.
20939 Turning off pre increment/decrement/modify allows IVOPTS to
20940 optimize DF/SF loops where the index is an int.
20942 2018-01-29 Richard Biener <rguenther@suse.de>
20943 Kelvin Nilsen <kelvin@gcc.gnu.org>
20946 * tree-vect-stmts.c (vectorizable_call): Don't call
20947 targetm.vectorize_builtin_md_vectorized_function if callee is
20950 2018-01-22 Carl Love <cel@us.ibm.com>
20952 * doc/extend.tex: Fix typo in second arg in
20953 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20955 2018-01-29 Richard Biener <rguenther@suse.de>
20957 PR tree-optimization/84086
20958 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20959 (flush_ssaname_freelist): When SSA names were released reset
20960 the SCEV hash table.
20962 2018-01-29 Richard Biener <rguenther@suse.de>
20964 PR tree-optimization/84057
20965 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20966 removed paths when removing edges.
20968 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20970 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20971 -mfunction-return=@var{choice}.
20973 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20975 PR diagnostic/84034
20976 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20977 Handle CR like TAB.
20978 (layout::print_source_line): Likewise.
20979 (test_get_line_width_without_trailing_whitespace): Add test cases.
20981 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20983 PR middle-end/84040
20984 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20987 2018-01-26 Jim Wilson <jimw@sifive.com>
20989 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20991 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20994 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20996 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20997 and CMP + SUB-immediate -> SUBS.
20999 2018-01-26 Martin Sebor <msebor@redhat.com>
21001 PR tree-optimization/83896
21002 * tree-ssa-strlen.c (get_string_len): Rename...
21003 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
21004 Avoid assuming length is constant.
21005 (handle_char_store): Use HOST_WIDE_INT for string length.
21007 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
21010 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
21011 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
21013 2018-01-26 Richard Biener <rguenther@suse.de>
21015 PR rtl-optimization/84003
21016 * dse.c (record_store): Only record redundant stores when
21017 the earlier store aliases at least all accesses the later one does.
21019 2018-01-26 Jakub Jelinek <jakub@redhat.com>
21021 PR rtl-optimization/83985
21022 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
21023 REG_CFA_RESTORE insns.
21024 (delete_unmarked_insns): Don't ignore separate shrink wrapping
21025 REG_CFA_RESTORE insns here.
21028 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
21029 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
21031 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21033 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
21034 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
21035 (arc_init): Likewise.
21036 (arc_override_options): Likewise.
21037 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
21039 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
21041 * config/arc/arc.h (TARGET_DBNZ): Define.
21042 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
21043 properly set the tune attribute.
21044 (dbnz): Use TARGET_DBNZ guard.
21045 * config/arc/arc.opt (mtune): Add core3 option.
21047 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21049 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
21050 recognize new pic like addresses.
21051 (arc_delegitimize_address): Clean up.
21053 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21055 * config/arc/arc-arches.def: Option mrf16 valid for all
21057 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
21058 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
21059 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
21060 * config/arc/arc-tables.opt: Regenerate.
21061 * config/arc/arc.c (arc_conditional_register_usage): Handle
21062 reduced register file case.
21063 (arc_file_start): Set must have build attributes.
21064 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
21065 mrf16 option value.
21066 * config/arc/arc.opt (mrf16): Add new option.
21067 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
21068 * config/arc/genmultilib.awk: Handle new mrf16 option.
21069 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
21070 * config/arc/t-multilib: Regenerate.
21071 * doc/invoke.texi (ARC Options): Document mrf16 option.
21073 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21075 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
21076 * config/arc/arc.c (arc_handle_secure_attribute): New function.
21077 (arc_attribute_table): Add 'secure_call' attribute.
21078 (arc_print_operand): Print secure call operand.
21079 (arc_function_ok_for_sibcall): Don't optimize tail calls when
21081 (arc_is_secure_call_p): New function. * config/arc/arc.md
21082 (call_i): Add support for sjli instruction.
21083 (call_value_i): Likewise.
21084 * config/arc/constraints.md (Csc): New constraint.
21086 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21087 John Eric Martin <John.Martin@emmicro-us.com>
21089 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
21090 * config/arc/arc.c (_arc_jli_section): New struct.
21091 (arc_jli_section): New type.
21092 (rc_jli_sections): New static variable.
21093 (arc_handle_jli_attribute): New function.
21094 (arc_attribute_table): Add jli_always and jli_fixed attribute.
21095 (arc_file_end): New function.
21096 (TARGET_ASM_FILE_END): Define.
21097 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
21098 (arc_add_jli_section): New function.
21099 (jli_call_scan): Likewise.
21100 (arc_reorg): Call jli_call_scan.
21101 (arc_output_addsi): Remove 'S' from printing asm operand.
21102 (arc_is_jli_call_p): New function.
21103 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
21105 (movhi_insn): Likewise.
21106 (movsi_insn): Likewise.
21107 (movsi_set_cc_insn): Likewise.
21108 (loadqi_update): Likewise.
21109 (load_zeroextendqisi_update): Likewise.
21110 (load_signextendqisi_update): Likewise.
21111 (loadhi_update): Likewise.
21112 (load_zeroextendhisi_update): Likewise.
21113 (load_signextendhisi_update): Likewise.
21114 (loadsi_update): Likewise.
21115 (loadsf_update): Likewise.
21116 (movsicc_insn): Likewise.
21117 (bset_insn): Likewise.
21118 (bxor_insn): Likewise.
21119 (bclr_insn): Likewise.
21120 (bmsk_insn): Likewise.
21121 (bicsi3_insn): Likewise.
21122 (cmpsi_cc_c_insn): Likewise.
21123 (movsi_ne): Likewise.
21124 (movsi_cond_exec): Likewise.
21125 (clrsbsi2): Likewise.
21126 (norm_f): Likewise.
21136 (call_i): Remove 'S' asm letter, add jli instruction.
21137 (call_value_i): Likewise.
21138 * config/arc/arc.op (mjli-always): New option.
21139 * config/arc/constraints.md (Cji): New constraint.
21140 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
21142 (subsf3_fpx): Likewise.
21143 (mulsf3_fpx): Likewise.
21144 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
21146 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
21147 function attrbutes.
21148 * doc/invoke.texi (ARC): Document mjli-always option.
21150 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
21152 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
21153 avoid addition with 0 and use incw and decw where possible.
21155 2018-01-26 Richard Biener <rguenther@suse.de>
21157 PR tree-optimization/81082
21158 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
21159 association if it requires casting to unsigned.
21160 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
21161 from fold_plusminus_mult_expr to catch important cases late when
21162 range info is available.
21164 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21166 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
21167 * configure.ac (hidden_linkonce): New test.
21168 * configure: Regenerate.
21169 * config.in: Regenerate.
21171 2018-01-26 Julia Koval <julia.koval@intel.com>
21173 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
21174 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
21175 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
21176 _mm_mask_bitshuffle_epi64_mask): Fix type.
21177 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
21178 USI_FTYPE_V4DI_V4DI_USI): Remove.
21179 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
21180 __builtin_ia32_vpshufbitqmb256_mask,
21181 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
21182 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
21183 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
21185 2018-01-26 Alan Modra <amodra@gmail.com>
21188 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
21189 UNSPEC_VBPERMQ. Sort other unspecs.
21191 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
21193 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
21195 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
21197 PR middle-end/83055
21198 * predict.c (drop_profile): Do not push/pop cfun; update also
21200 (handle_missing_profiles): Fix logic looking for zero profiles.
21202 2018-01-25 Jakub Jelinek <jakub@redhat.com>
21204 PR middle-end/83977
21205 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
21206 on functions with #pragma omp declare simd or functions with simd
21208 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
21209 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
21210 Remove trailing \n from warning_at calls.
21212 2018-01-25 Tom de Vries <tom@codesourcery.com>
21215 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21216 for neutered workers.
21218 2018-01-24 Joseph Myers <joseph@codesourcery.com>
21221 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
21222 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
21224 2018-01-24 Jeff Law <law@redhat.com>
21227 * i386.c (get_probe_interval): Move to earlier point.
21228 (ix86_compute_frame_layout): If -fstack-clash-protection and
21229 the frame is larger than the probe interval, then use pushes
21230 to save registers rather than reg->mem moves.
21231 (ix86_expand_prologue): Remove conditional for int_registers_saved
21234 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
21237 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
21238 min/max for never referenced object.
21240 2018-01-24 Jakub Jelinek <jakub@redhat.com>
21242 PR middle-end/83977
21243 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
21245 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
21246 attributes from DECL_ATTRIBUTES (decl) without affecting
21247 DECL_ATTRIBUTES (current_function_decl).
21248 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
21249 functions with non-NULL DECL_ABSTRACT_ORIGIN.
21251 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
21253 PR tree-optimization/83979
21254 * fold-const.c (fold_comparison): Use constant_boolean_node
21255 instead of boolean_{true,false}_node.
21257 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
21259 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
21262 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21264 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21265 Simplify the clause that sets the length attribute.
21266 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21267 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
21268 clause that sets the length attribute.
21269 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21271 2018-01-24 Tom de Vries <tom@codesourcery.com>
21274 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
21275 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
21276 Add strict parameter.
21277 (prevent_branch_around_nothing): Insert dummy insn between branch to
21278 label and label with no ptx insn inbetween.
21279 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
21281 2018-01-24 Tom de Vries <tom@codesourcery.com>
21284 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21285 for neutered threads in warp.
21286 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
21288 2018-01-24 Richard Biener <rguenther@suse.de>
21290 PR tree-optimization/83176
21291 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
21294 2018-01-24 Richard Biener <rguenther@suse.de>
21296 PR tree-optimization/82819
21297 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
21298 code generating pluses that are no-ops in the target precision.
21300 2018-01-24 Richard Biener <rguenther@suse.de>
21302 PR middle-end/84000
21303 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
21305 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21307 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
21308 to merge probabilities.
21309 * predict.c (probably_never_executed): Also mark as cold functions
21310 with global 0 profile and guessed local profile.
21311 * profile-count.c (profile_probability::combine_with_count): New
21313 * profile-count.h (profile_probability::operator*,
21314 profile_probability::operator*=, profile_probability::operator/,
21315 profile_probability::operator/=): Reduce precision to adjusted
21316 and set value to guessed on contradictory divisions.
21317 (profile_probability::combine_with_freq): Remove.
21318 (profile_probability::combine_wiht_count): Declare.
21319 (profile_count::force_nonzero):: Set to adjusted.
21320 (profile_count::probability_in):: Set quality to adjusted.
21321 * tree-ssa-tail-merge.c (replace_block_by): Use
21322 combine_with_count.
21324 2018-01-23 Andrew Waterman <andrew@sifive.com>
21325 Jim Wilson <jimw@sifive.com>
21327 * config/riscv/riscv.c (riscv_stack_boundary): New.
21328 (riscv_option_override): Set riscv_stack_boundary. Handle
21329 riscv_preferred_stack_boundary_arg.
21330 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
21331 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
21332 (STACK_BOUNDARY): Set to riscv_stack_boundary.
21333 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
21334 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
21335 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
21337 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
21340 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
21341 of struct ix86_frame.
21342 (ix86_expand_epilogue): Likewise. Add a local variable for
21343 the reg_save_offset field in struct ix86_frame.
21345 2018-01-23 Bin Cheng <bin.cheng@arm.com>
21347 PR tree-optimization/82604
21348 * tree-loop-distribution.c (enum partition_kind): New enum item
21349 PKIND_PARTIAL_MEMSET.
21350 (partition_builtin_p): Support above new enum item.
21351 (generate_code_for_partition): Ditto.
21352 (compute_access_range): Differentiate cases that equality can be
21353 proven at all loops, the innermost loops or no loops.
21354 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
21355 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
21356 (finalize_partitions, distribute_loop): Don't fuse partition of
21357 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
21358 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
21359 parloop is enabled.
21361 2018-01-23 Martin Liska <mliska@suse.cz>
21363 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
21364 order to ignore the predictor.
21365 (PRED_POLYMORPHIC_CALL): Likewise.
21366 (PRED_RECURSIVE_CALL): Likewise.
21368 2018-01-23 Martin Liska <mliska@suse.cz>
21370 * tree-profile.c (tree_profiling): Print function header to
21371 aware reader which function we are working on.
21372 * value-prof.c (gimple_find_values_to_profile): Do not print
21373 not interesting value histograms.
21375 2018-01-23 Martin Liska <mliska@suse.cz>
21377 * profile-count.h (enum profile_quality): Add
21378 profile_uninitialized as the first value. Do not number values
21379 as they are zero based.
21380 (profile_count::verify): Update sanity check.
21381 (profile_probability::verify): Likewise.
21383 2018-01-23 Nathan Sidwell <nathan@acm.org>
21385 * doc/invoke.texi (ffor-scope): Deprecate.
21387 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21389 PR tree-optimization/83510
21390 * domwalk.c (set_all_edges_as_executable): New function.
21391 (dom_walker::dom_walker): Convert bool param
21392 "skip_unreachable_blocks" to enum reachability. Move setup of
21393 edge flags to set_all_edges_as_executable and only do it when
21394 reachability is REACHABLE_BLOCKS.
21395 * domwalk.h (enum dom_walker::reachability): New enum.
21396 (dom_walker::dom_walker): Convert bool param
21397 "skip_unreachable_blocks" to enum reachability.
21398 (set_all_edges_as_executable): New decl.
21399 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
21400 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
21402 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
21403 but converting true to REACHABLE_BLOCKS.
21404 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
21406 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
21407 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
21408 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
21410 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
21411 if check_all_array_refs will be called.
21413 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21415 * tree.c (selftest::test_location_wrappers): Add more test
21418 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21420 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
21421 (selftest::test_bit_in_range): Likewise.
21423 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
21426 * doc/sourcebuild.texi (vect_float): Say that the selector
21427 only describes the situation when -funsafe-math-optimizations is on.
21428 (vect_float_strict): Document.
21430 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
21432 PR tree-optimization/83965
21433 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
21434 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
21435 instead of checking only for a reduction.
21436 (vect_recog_widen_sum_pattern): Likewise.
21438 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21440 * predict.c (probably_never_executed): Only use precise profile info.
21441 (compute_function_frequency): Skip after inlining hack since we now
21442 have quality checking.
21444 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21446 * profile-count.h (profile_probability::very_unlikely,
21447 profile_probability::unlikely, profile_probability::even): Set
21448 precision to guessed.
21450 2018-01-23 Richard Biener <rguenther@suse.de>
21452 PR tree-optimization/83963
21453 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21454 Properly terminate dominator walk when crossing the exit edge not
21455 when visiting its source block.
21457 2018-01-23 Jakub Jelinek <jakub@redhat.com>
21460 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21461 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21463 2018-01-22 Jakub Jelinek <jakub@redhat.com>
21465 PR tree-optimization/83957
21466 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
21467 semicolon after for body surrounded by braces.
21469 PR tree-optimization/83081
21470 * profile-count.h (profile_probability::split): New method.
21471 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21472 Use profile_probability::split.
21473 (do_compare_rtx_and_jump): Fix adjustment of probabilities
21474 when splitting a single conditional jump into 2.
21476 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21478 PR tree-optimization/69452
21479 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21482 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21484 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21485 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21486 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21488 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21490 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21492 * config/rl78/rl78.md (movdi): New define_expand.
21493 * config/rl78/rl78.c (rl78_split_movdi): New function.
21495 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21498 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21500 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21501 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21502 128-bit to produce an UNSPEC move to get the double word with the
21503 signbit and then a shift directly to do signbit.
21504 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21505 implementation with a new version that just does either a direct
21506 move or a regular move. Move memory interface to separate insns.
21507 Move insns so they are next to the expander.
21508 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21509 with signbit move. Split big and little endian case.
21510 (signbit<mode>2_dm_mem_le): Likewise.
21511 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21512 (signbit<mode>2_dm2): Likewise.
21514 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21516 * config/rl78/rl78.md (anddi3): New define_expand.
21518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21520 * config/rl78/rl78.md (umindi3): New define_expand.
21522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21524 * config/rl78/rl78.md (smindi3): New define_expand.
21526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21528 * config/rl78/rl78.md (smaxdi3): New define_expand.
21530 2018-01-22 Carl Love <cel@us.ibm.com>
21532 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21533 LVX_V1TI): Add macro expansion.
21534 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21535 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21536 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21537 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21538 Change check to determine if the instruction is a byte reversing
21539 entry. Fix typo in comment.
21540 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21541 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21542 Add def_builtin calls for new builtins.
21543 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21544 Add define_insn expansion.
21546 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21548 * config/rl78/rl78.md (umaxdi3): New define_expand.
21550 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21552 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21553 for non-QImode registers.
21555 2018-01-22 Richard Biener <rguenther@suse.de>
21557 PR tree-optimization/83963
21558 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21559 including the loop exit block.
21560 (scop_detection::merge_sese): Likewise.
21561 (scop_detection::add_scop): Do it here instead.
21563 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21565 * doc/sourcebuild.texi (arm_softfloat): Document.
21567 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21570 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21571 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21572 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21574 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21575 David Edelsohn <dje.gcc@gmail.com>
21578 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21579 Change "crset eq" to "crset 2".
21580 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21581 (*call_indirect_aix<mode>_nospec): Likewise.
21582 (*call_value_indirect_aix<mode>_nospec): Likewise.
21583 (*call_indirect_elfv2<mode>_nospec): Likewise.
21584 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21585 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21586 change assembly output from . to $.
21587 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21588 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21589 (*tablejump<mode>_internal1_nospec): Likewise.
21591 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21594 * config/sh/sh_optimize_sett_clrt.cc:
21595 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21597 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21599 PR tree-optimization/83940
21600 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21601 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21602 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21603 use it to pass back the definition type.
21604 (vect_check_store_rhs): Likewise rhs_dt_out.
21605 (vect_build_gather_load_calls): Add a mask_dt argument and use
21606 it instead of a call to vect_is_simple_use.
21607 (vectorizable_store): Update calls to vect_check_load_store_mask
21608 and vect_check_store_rhs. Use the dt returned by the latter instead
21609 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21610 instead of calls to vect_is_simple_use. Pass the scalar rather
21611 than the vector operand to vect_is_simple_use when handling
21612 second and subsequent copies of an rhs value.
21613 (vectorizable_load): Update calls to vect_check_load_store_mask
21614 and vect_build_gather_load_calls. Use the cached mask_dt and
21615 gs_info.offset_dt instead of calls to vect_is_simple_use.
21617 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21619 PR middle-end/83945
21620 * tree-emutls.c: Include gimplify.h.
21621 (lower_emutls_2): New function.
21622 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21623 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21624 it before further processing.
21627 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21628 UINTVAL (trueop1) instead of INTVAL (op1).
21630 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21634 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21635 INCOMING_FRAME_SP_OFFSET if not defined.
21636 (scan_trace): Add ENTRY argument. If true and
21637 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21638 emit a note to adjust the CFA offset.
21639 (create_cfi_notes): Adjust scan_trace callers.
21640 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21641 INCOMING_FRAME_SP_OFFSET in the CIE.
21642 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21643 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21645 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21646 * doc/tm.texi: Regenerated.
21648 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21650 PR rtl-optimization/83147
21651 * lra-constraints.c (remove_inheritance_pseudos): Use
21652 lra_substitute_pseudo_within_insn.
21654 2018-01-19 Tom de Vries <tom@codesourcery.com>
21655 Cesar Philippidis <cesar@codesourcery.com>
21658 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21660 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21663 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21664 spaces for function labels.
21666 2018-01-19 Martin Liska <mliska@suse.cz>
21668 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21669 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21670 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21671 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21672 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21673 (PRED_CONST_RETURN): Change from 69 to 65.
21674 (PRED_NULL_RETURN): Change from 91 to 71.
21675 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21676 (PRED_LOOP_GUARD): Change from 66 to 73.
21678 2018-01-19 Martin Liska <mliska@suse.cz>
21680 * predict.c (predict_insn_def): Add new assert.
21681 (struct branch_predictor): Change type to signed integer.
21682 (test_prediction_value_range): Amend test to cover
21683 PROB_UNINITIALIZED.
21684 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21685 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21686 (PRED_LOOP_ITERATIONS_MAX): Likewise.
21687 (PRED_LOOP_IV_COMPARE): Likewise.
21688 * predict.h (PROB_UNINITIALIZED): Define new constant.
21690 2018-01-19 Martin Liska <mliska@suse.cz>
21692 * predict.c (dump_prediction): Add new format for
21693 analyze_brprob.py script which is enabled with -details
21695 * profile-count.h (precise_p): New function.
21697 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21699 PR tree-optimization/83922
21700 * tree-vect-loop.c (vect_verify_full_masking): Return false if
21701 there are no statements that need masking.
21702 (vect_active_double_reduction_p): New function.
21703 (vect_analyze_loop_operations): Use it when handling phis that
21704 are not in the loop header.
21706 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21708 PR tree-optimization/83914
21709 * tree-vect-loop.c (vectorizable_induction): Don't convert
21710 init_expr or apply the peeling adjustment for inductions
21711 that are nested within the vectorized loop.
21713 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21715 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21718 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21722 * function.h (gimplify_parameters): Add gimple_seq * argument.
21723 * function.c: Include gimple.h and options.h.
21724 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21725 for the added local temporaries if needed.
21726 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21727 if there are any parameter cleanups, wrap whole body into a
21728 try/finally with the cleanups.
21730 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
21733 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21734 Use GET_MODE_CLASS for scalar floating point.
21736 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21740 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21741 Fix call of call_cgraph_insertion_hooks.
21743 2018-01-18 Martin Sebor <msebor@redhat.com>
21745 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21747 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21750 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21753 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
21756 * common.opt: (-ffile-prefix-map): New option.
21757 * opts.c (common_handle_option): Defer it.
21758 * opts-global.c (handle_common_deferred_options): Handle it.
21759 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21760 * file-prefix-map.h: New file.
21761 (remap_debug_filename, add_debug_prefix_map): ...here.
21762 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21763 * final.c (debug_prefix_map, add_debug_prefix_map
21764 remap_debug_filename): Move to...
21765 * file-prefix-map.c: New file.
21766 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21767 generalize, get rid of alloca(), use strrchr() instead of strchr().
21768 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21769 Implement in terms of add_prefix_map().
21770 (remap_macro_filename, remap_debug_filename): Implement in term of
21772 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21773 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21774 * dbxout.c: Include file-prefix-map.h.
21775 * varasm.c: Likewise.
21776 * vmsdbgout.c: Likewise.
21777 * xcoffout.c: Likewise.
21778 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21779 * doc/cppopts.texi (-fmacro-prefix-map): Document.
21780 * doc/invoke.texi (-ffile-prefix-map): Document.
21781 (-fdebug-prefix-map): Update description.
21783 2018-01-18 Martin Liska <mliska@suse.cz>
21785 * config/i386/i386.c (indirect_thunk_name): Document that also
21787 (output_indirect_thunk): Document why both instructions
21788 (pause and lfence) are generated.
21790 2018-01-18 Richard Biener <rguenther@suse.de>
21792 PR tree-optimization/83887
21793 * graphite-scop-detection.c
21794 (scop_detection::get_nearest_dom_with_single_entry): Remove.
21795 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21796 (scop_detection::merge_sese): Re-implement with a flood-fill
21797 algorithm that properly finds a SESE region if it exists.
21799 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21802 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21803 pointer_diff optimizations use view_convert instead of convert.
21805 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21807 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21808 Generate different code for -mno-speculate-indirect-jumps.
21809 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21810 (*call_indirect_aix<mode>): Disable for
21811 -mno-speculate-indirect-jumps.
21812 (*call_indirect_aix<mode>_nospec): New define_insn.
21813 (*call_value_indirect_aix<mode>): Disable for
21814 -mno-speculate-indirect-jumps.
21815 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21816 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21817 -mno-speculate-indirect-jumps.
21818 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21820 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21822 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21823 long double type, set the flags for noting the default long double
21824 type, even if we don't pass or return a long double type.
21826 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21829 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21832 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21834 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21835 support for merge[hl].
21836 (fold_mergehl_helper): New helper function.
21837 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21838 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21839 (altivec_vmrglw_direct): Add xxmrglw insn.
21841 2018-01-17 Andrew Waterman <andrew@sifive.com>
21843 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21844 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21846 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21849 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21850 call the lto_location_cache before reading the
21851 DECL_SOURCE_LOCATION of the types.
21853 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21854 Richard Sandiford <richard.sandiford@linaro.org>
21856 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21857 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21858 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21859 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21860 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21862 * config/aarch64/constraints.md (aarch64_movti_operand):
21864 * config/aarch64/predicates.md (Uti): Add new constraint.
21866 2018-01-17 Carl Love <cel@us.ibm.com>
21868 * config/rs6000/vsx.md (define_expand xl_len_r,
21869 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21870 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21872 (define_expand, define_insn): Move the shift left from the
21873 define_insn to the define_expand for lxvl and stxvl instructions.
21874 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21875 and XL_LEN_R definitions to PURE.
21877 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21879 * config/i386/i386.c (indirect_thunk_name): Declare regno
21880 as unsigned int. Compare regno with INVALID_REGNUM.
21881 (output_indirect_thunk): Ditto.
21882 (output_indirect_thunk_function): Ditto.
21883 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21884 in the call to output_indirect_thunk_function.
21886 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21888 PR middle-end/83884
21889 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21890 rather than the size of inner_type to determine the stack slot size
21891 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21893 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21896 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21899 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21901 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21902 endian Linux systems to optionally enable multilibs for selecting
21903 the long double type if the user configured an explicit type.
21904 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21905 have no long double multilibs if not defined.
21906 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21907 warn if the user used -mabi={ieee,ibm}longdouble and we built
21908 multilibs for long double.
21909 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21910 appropriate multilib option.
21911 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21913 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21914 for building long double multilibs.
21915 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21917 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21919 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21922 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21924 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21927 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21930 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21933 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21935 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21936 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21938 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21940 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21941 different rtl trees depending on TARGET_64BIT.
21942 (rs6000_gen_lvx): Likewise.
21944 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21946 * config/visium/visium.md (nop): Tweak comment.
21947 (hazard_nop): Likewise.
21949 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21951 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21952 -mspeculate-indirect-jumps.
21953 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21954 for -mno-speculate-indirect-jumps.
21955 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21956 (*call_value_indirect_elfv2<mode>): Disable for
21957 -mno-speculate-indirect-jumps.
21958 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21959 (indirect_jump): Emit different RTL for
21960 -mno-speculate-indirect-jumps.
21961 (*indirect_jump<mode>): Disable for
21962 -mno-speculate-indirect-jumps.
21963 (*indirect_jump<mode>_nospec): New define_insn.
21964 (tablejump): Emit different RTL for
21965 -mno-speculate-indirect-jumps.
21966 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21967 (tablejumpsi_nospec): New define_expand.
21968 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21969 (tablejumpdi_nospec): New define_expand.
21970 (*tablejump<mode>_internal1): Disable for
21971 -mno-speculate-indirect-jumps.
21972 (*tablejump<mode>_internal1_nospec): New define_insn.
21973 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21976 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21978 * caller-save.c (insert_save): Drop unnecessary parameter. All
21981 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21982 Richard Biener <rguenth@suse.de>
21985 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21986 return early, inline manually is_gimple_sizepos. Make sure if we
21987 call gimplify_expr we don't end up with a gimple constant.
21988 * tree.c (variably_modified_type_p): Don't return true for
21989 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21990 * gimplify.h (is_gimple_sizepos): Remove.
21992 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21994 PR tree-optimization/83857
21995 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21996 vectorizable_live_operation for pure SLP statements.
21997 (vectorizable_live_operation): Handle PHIs.
21999 2018-01-16 Richard Biener <rguenther@suse.de>
22001 PR tree-optimization/83867
22002 * tree-vect-stmts.c (vect_transform_stmt): Precompute
22003 nested_in_vect_loop_p since the scalar stmt may get invalidated.
22005 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22008 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
22009 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
22010 If off is not INTEGER_CST, issue a may not be aligned warning
22011 rather than isn't aligned. Use isn%'t rather than isn't.
22012 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
22014 <case MULT_EXPR>: Improve the case when bottom and one of the
22015 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
22016 operand, in that case check if the other operand is multiple of
22017 bottom divided by the INTEGER_CST operand.
22019 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22022 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
22023 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
22024 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
22025 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
22026 * config/pa/pa.c (pa_function_arg_advance): Likewise.
22027 (pa_function_arg, pa_arg_partial_bytes): Likewise.
22028 (pa_function_arg_size): New function.
22030 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22032 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
22033 in a separate statement.
22035 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22037 PR tree-optimization/83847
22038 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
22039 group gathers and scatters.
22041 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22043 PR rtl-optimization/86620
22044 * params.def (max-sched-ready-insns): Bump minimum value to 1.
22046 PR rtl-optimization/83213
22047 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
22048 to last if both are JUMP_INSNs.
22050 PR tree-optimization/83843
22051 * gimple-ssa-store-merging.c
22052 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
22053 store_immediate_info for bswap/nop orig_stores.
22055 2018-01-15 Andrew Waterman <andrew@sifive.com>
22057 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
22059 <UDIV>: Increase cost if !TARGET_DIV.
22061 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
22063 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
22064 (define_attr "cr_logical_3op"): New.
22065 (cceq_ior_compare): Adjust.
22066 (cceq_ior_compare_complement): Adjust.
22067 (*cceq_rev_compare): Adjust.
22068 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
22069 (is_cracked_insn): Adjust.
22070 (insn_must_be_first_in_group): Adjust.
22071 * config/rs6000/40x.md: Adjust.
22072 * config/rs6000/440.md: Adjust.
22073 * config/rs6000/476.md: Adjust.
22074 * config/rs6000/601.md: Adjust.
22075 * config/rs6000/603.md: Adjust.
22076 * config/rs6000/6xx.md: Adjust.
22077 * config/rs6000/7450.md: Adjust.
22078 * config/rs6000/7xx.md: Adjust.
22079 * config/rs6000/8540.md: Adjust.
22080 * config/rs6000/cell.md: Adjust.
22081 * config/rs6000/e300c2c3.md: Adjust.
22082 * config/rs6000/e500mc.md: Adjust.
22083 * config/rs6000/e500mc64.md: Adjust.
22084 * config/rs6000/e5500.md: Adjust.
22085 * config/rs6000/e6500.md: Adjust.
22086 * config/rs6000/mpc.md: Adjust.
22087 * config/rs6000/power4.md: Adjust.
22088 * config/rs6000/power5.md: Adjust.
22089 * config/rs6000/power6.md: Adjust.
22090 * config/rs6000/power7.md: Adjust.
22091 * config/rs6000/power8.md: Adjust.
22092 * config/rs6000/power9.md: Adjust.
22093 * config/rs6000/rs64.md: Adjust.
22094 * config/rs6000/titan.md: Adjust.
22096 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22098 * config/i386/predicates.md (indirect_branch_operand): Rewrite
22099 ix86_indirect_branch_register logic.
22101 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22103 * config/i386/constraints.md (Bs): Update
22104 ix86_indirect_branch_register check. Don't check
22105 ix86_indirect_branch_register with GOT_memory_operand.
22107 * config/i386/predicates.md (GOT_memory_operand): Don't check
22108 ix86_indirect_branch_register here.
22109 (GOT32_symbol_operand): Likewise.
22111 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22113 * config/i386/predicates.md (constant_call_address_operand):
22114 Rewrite ix86_indirect_branch_register logic.
22115 (sibcall_insn_operand): Likewise.
22117 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22119 * config/i386/constraints.md (Bs): Replace
22120 ix86_indirect_branch_thunk_register with
22121 ix86_indirect_branch_register.
22123 * config/i386/i386.md (indirect_jump): Likewise.
22124 (tablejump): Likewise.
22125 (*sibcall_memory): Likewise.
22126 (*sibcall_value_memory): Likewise.
22127 Peepholes of indirect call and jump via memory: Likewise.
22128 * config/i386/i386.opt: Likewise.
22129 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22130 (GOT_memory_operand): Likewise.
22131 (call_insn_operand): Likewise.
22132 (sibcall_insn_operand): Likewise.
22133 (GOT32_symbol_operand): Likewise.
22135 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22137 PR middle-end/83837
22138 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
22139 type rather than type addr's type points to.
22140 (expand_omp_atomic_mutex): Likewise.
22141 (expand_omp_atomic): Likewise.
22143 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22146 * config/i386/i386.c (output_indirect_thunk_function): Use
22147 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
22148 for __x86_return_thunk.
22150 2018-01-15 Richard Biener <rguenther@suse.de>
22152 PR middle-end/83850
22153 * expmed.c (extract_bit_field_1): Fix typo.
22155 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22158 * config/arm/iterators.md (VF): New mode iterator.
22159 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
22160 Remove integer-related logic from pattern.
22161 (neon_vabd<mode>_3): Likewise.
22163 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22165 PR middle-end/82694
22166 * common.opt (fstrict-overflow): No longer an alias.
22167 (fwrapv-pointer): New option.
22168 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
22169 also for pointer types based on flag_wrapv_pointer.
22170 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
22171 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
22172 opts->x_flag_wrapv got set.
22173 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
22174 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
22175 POINTER_TYPE_OVERFLOW_UNDEFINED.
22176 * match.pd: Likewise in address comparison pattern.
22177 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
22179 2018-01-15 Richard Biener <rguenther@suse.de>
22182 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
22183 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
22184 Reset type names to their identifier if their TYPE_DECL doesn't
22185 have linkage (and thus is used for ODR and devirt).
22186 (save_debug_info_for_decl): Remove.
22187 (save_debug_info_for_type): Likewise.
22188 (add_tree_to_fld_list): Adjust.
22189 * tree-pretty-print.c (dump_generic_node): Make dumping of
22190 type names more robust.
22192 2018-01-15 Richard Biener <rguenther@suse.de>
22194 * BASE-VER: Bump to 8.0.1.
22196 2018-01-14 Martin Sebor <msebor@redhat.com>
22199 * builtins.c (check_access): Avoid warning when the no-warning bit
22202 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
22204 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
22205 * ira-color (allocno_hard_regs_compare): Likewise.
22207 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
22210 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
22211 Use .pushsection/.popsection.
22213 2018-01-14 Martin Sebor <msebor@redhat.com>
22216 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
22218 2018-01-14 Jakub Jelinek <jakub@redhat.com>
22220 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
22221 entry from extra_headers.
22222 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
22223 extra_headers, make the list bitwise identical to the i?86-*-* one.
22225 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22227 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
22228 -mcmodel=large with -mindirect-branch=thunk,
22229 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
22230 -mfunction-return=thunk-extern.
22231 * doc/invoke.texi: Document -mcmodel=large is incompatible with
22232 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
22233 -mfunction-return=thunk and -mfunction-return=thunk-extern.
22235 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22237 * config/i386/i386.c (print_reg): Print the name of the full
22238 integer register without '%'.
22239 (ix86_print_operand): Handle 'V'.
22240 * doc/extend.texi: Document 'V' modifier.
22242 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22244 * config/i386/constraints.md (Bs): Disallow memory operand for
22245 -mindirect-branch-register.
22247 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22248 (GOT_memory_operand): Likewise.
22249 (call_insn_operand): Likewise.
22250 (sibcall_insn_operand): Likewise.
22251 (GOT32_symbol_operand): Likewise.
22252 * config/i386/i386.md (indirect_jump): Call convert_memory_address
22253 for -mindirect-branch-register.
22254 (tablejump): Likewise.
22255 (*sibcall_memory): Likewise.
22256 (*sibcall_value_memory): Likewise.
22257 Disallow peepholes of indirect call and jump via memory for
22258 -mindirect-branch-register.
22259 (*call_pop): Replace m with Bw.
22260 (*call_value_pop): Likewise.
22261 (*sibcall_pop_memory): Replace m with Bs.
22262 * config/i386/i386.opt (mindirect-branch-register): New option.
22263 * doc/invoke.texi: Document -mindirect-branch-register option.
22265 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22267 * config/i386/i386-protos.h (ix86_output_function_return): New.
22268 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
22269 set function_return_type.
22270 (indirect_thunk_name): Add ret_p to indicate thunk for function
22272 (output_indirect_thunk_function): Pass false to
22273 indirect_thunk_name.
22274 (ix86_output_indirect_branch_via_reg): Likewise.
22275 (ix86_output_indirect_branch_via_push): Likewise.
22276 (output_indirect_thunk_function): Create alias for function
22277 return thunk if regno < 0.
22278 (ix86_output_function_return): New function.
22279 (ix86_handle_fndecl_attribute): Handle function_return.
22280 (ix86_attribute_table): Add function_return.
22281 * config/i386/i386.h (machine_function): Add
22282 function_return_type.
22283 * config/i386/i386.md (simple_return_internal): Use
22284 ix86_output_function_return.
22285 (simple_return_internal_long): Likewise.
22286 * config/i386/i386.opt (mfunction-return=): New option.
22287 (indirect_branch): Mention -mfunction-return=.
22288 * doc/extend.texi: Document function_return function attribute.
22289 * doc/invoke.texi: Document -mfunction-return= option.
22291 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22293 * config/i386/i386-opts.h (indirect_branch): New.
22294 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
22295 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
22296 with local indirect jump when converting indirect call and jump.
22297 (ix86_set_indirect_branch_type): New.
22298 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
22299 (indirectlabelno): New.
22300 (indirect_thunk_needed): Likewise.
22301 (indirect_thunk_bnd_needed): Likewise.
22302 (indirect_thunks_used): Likewise.
22303 (indirect_thunks_bnd_used): Likewise.
22304 (INDIRECT_LABEL): Likewise.
22305 (indirect_thunk_name): Likewise.
22306 (output_indirect_thunk): Likewise.
22307 (output_indirect_thunk_function): Likewise.
22308 (ix86_output_indirect_branch_via_reg): Likewise.
22309 (ix86_output_indirect_branch_via_push): Likewise.
22310 (ix86_output_indirect_branch): Likewise.
22311 (ix86_output_indirect_jmp): Likewise.
22312 (ix86_code_end): Call output_indirect_thunk_function if needed.
22313 (ix86_output_call_insn): Call ix86_output_indirect_branch if
22315 (ix86_handle_fndecl_attribute): Handle indirect_branch.
22316 (ix86_attribute_table): Add indirect_branch.
22317 * config/i386/i386.h (machine_function): Add indirect_branch_type
22318 and has_local_indirect_jump.
22319 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
22321 (tablejump): Likewise.
22322 (*indirect_jump): Use ix86_output_indirect_jmp.
22323 (*tablejump_1): Likewise.
22324 (simple_return_indirect_internal): Likewise.
22325 * config/i386/i386.opt (mindirect-branch=): New option.
22326 (indirect_branch): New.
22329 (thunk-inline): Likewise.
22330 (thunk-extern): Likewise.
22331 * doc/extend.texi: Document indirect_branch function attribute.
22332 * doc/invoke.texi: Document -mindirect-branch= option.
22334 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
22337 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
22339 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
22341 * ipa-inline.c (want_inline_small_function_p): Return false if
22342 inlining has already failed with CIF_FINAL_ERROR.
22343 (update_caller_keys): Call want_inline_small_function_p before
22345 (update_callee_keys): Likewise.
22347 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
22349 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
22351 (rs6000_quadword_masked_address_p): Likewise.
22352 (quad_aligned_load_p): Likewise.
22353 (quad_aligned_store_p): Likewise.
22354 (const_load_sequence_p): Add comment to describe the outer-most loop.
22355 (mimic_memory_attributes_and_flags): New function.
22356 (rs6000_gen_stvx): Likewise.
22357 (replace_swapped_aligned_store): Likewise.
22358 (rs6000_gen_lvx): Likewise.
22359 (replace_swapped_aligned_load): Likewise.
22360 (replace_swapped_load_constant): Capitalize argument name in
22361 comment describing this function.
22362 (rs6000_analyze_swaps): Add a third pass to search for vector loads
22363 and stores that access quad-word aligned addresses and replace
22364 with stvx or lvx instructions when appropriate.
22365 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
22366 New function prototype.
22367 (rs6000_quadword_masked_address_p): Likewise.
22368 (rs6000_gen_lvx): Likewise.
22369 (rs6000_gen_stvx): Likewise.
22370 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
22371 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
22372 when memory address is aligned.
22373 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
22374 this split to select lvx instruction when memory address is aligned.
22375 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
22376 instruction when memory address is aligned.
22377 (*vsx_le_perm_load_v16qi): Likewise.
22378 (four unnamed splitters): Modify to select the stvx instruction
22379 when memory is aligned.
22381 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
22383 * predict.c (determine_unlikely_bbs): Handle correctly BBs
22384 which appears in the queue multiple times.
22386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22387 Alan Hayward <alan.hayward@arm.com>
22388 David Sherwood <david.sherwood@arm.com>
22390 * tree-vectorizer.h (vec_lower_bound): New structure.
22391 (_loop_vec_info): Add check_nonzero and lower_bounds.
22392 (LOOP_VINFO_CHECK_NONZERO): New macro.
22393 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
22394 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
22395 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
22396 fields. Make seg_len the distance travelled, not including the
22398 (dr_direction_indicator): Declare.
22399 (dr_zero_step_indicator): Likewise.
22400 (dr_known_forward_stride_p): Likewise.
22401 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
22403 (runtime_alias_check_p): Allow runtime alias checks with
22405 (operator ==): Compare access_size and align.
22406 (prune_runtime_alias_test_list): Rework for new distinction between
22407 the access_size and seg_len.
22408 (create_intersect_range_checks_index): Likewise. Cope with polynomial
22410 (get_segment_min_max): New function.
22411 (create_intersect_range_checks): Use it.
22412 (dr_step_indicator): New function.
22413 (dr_direction_indicator): Likewise.
22414 (dr_zero_step_indicator): Likewise.
22415 (dr_known_forward_stride_p): Likewise.
22416 * tree-loop-distribution.c (data_ref_segment_size): Return
22417 DR_STEP * (niters - 1).
22418 (compute_alias_check_pairs): Update call to the dr_with_seg_len
22420 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
22421 (vect_preserves_scalar_order_p): New function, split out from...
22422 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
22423 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
22424 (vect_vfa_access_size): New function.
22425 (vect_vfa_align): Likewise.
22426 (vect_compile_time_alias): Take access_size_a and access_b arguments.
22427 (dump_lower_bound): New function.
22428 (vect_check_lower_bound): Likewise.
22429 (vect_small_gap_p): Likewise.
22430 (vectorizable_with_step_bound_p): Likewise.
22431 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
22432 depencies if the vectorization factor is 1. Convert the checks
22433 for nonzero steps into checks on the bounds of DR_STEP. Try using
22434 a bunds check for variable steps if the minimum required step is
22435 relatively small. Update calls to the dr_with_seg_len
22436 constructor and to vect_compile_time_alias.
22437 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
22439 (vect_loop_versioning): Call it.
22440 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
22442 (vect_estimate_min_profitable_iters): Account for any bounds checks.
22444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22445 Alan Hayward <alan.hayward@arm.com>
22446 David Sherwood <david.sherwood@arm.com>
22448 * doc/sourcebuild.texi (vect_scatter_store): Document.
22449 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22451 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22453 * genopinit.c (main): Add supports_vec_scatter_store and
22454 supports_vec_scatter_store_cached to target_optabs.
22455 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22456 IFN_MASK_SCATTER_STORE.
22457 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22459 * internal-fn.h (internal_store_fn_p): Declare.
22460 (internal_fn_stored_value_index): Likewise.
22461 * internal-fn.c (scatter_store_direct): New macro.
22462 (expand_scatter_store_optab_fn): New function.
22463 (direct_scatter_store_optab_supported_p): New macro.
22464 (internal_store_fn_p): New function.
22465 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22466 IFN_MASK_SCATTER_STORE.
22467 (internal_fn_mask_index): Likewise.
22468 (internal_fn_stored_value_index): New function.
22469 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22470 for scatter stores.
22471 * optabs-query.h (supports_vec_scatter_store_p): Declare.
22472 * optabs-query.c (supports_vec_scatter_store_p): New function.
22473 * tree-vectorizer.h (vect_get_store_rhs): Declare.
22474 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22475 true for scatter stores.
22476 (vect_gather_scatter_fn_p): Handle scatter stores too.
22477 (vect_check_gather_scatter): Consider using scatter stores if
22478 supports_vec_scatter_store_p.
22479 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22480 scatter stores too.
22481 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22482 internal_fn_stored_value_index.
22483 (check_load_store_masking): Handle scatter stores too.
22484 (vect_get_store_rhs): Make public.
22485 (vectorizable_call): Use internal_store_fn_p.
22486 (vectorizable_store): Handle scatter store internal functions.
22487 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22488 when deciding whether the end of the group has been reached.
22489 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22490 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22491 (mask_scatter_store<mode>): New insns.
22493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22494 Alan Hayward <alan.hayward@arm.com>
22495 David Sherwood <david.sherwood@arm.com>
22497 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22498 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22499 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22501 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22502 Use vect_truncate_gather_scatter_offset if we can't treat the
22503 operation as a normal gather load or scatter store.
22504 (get_group_load_store_type): Take the gather_scatter_info
22505 as argument. Try using a gather load or scatter store for
22506 single-element groups.
22507 (get_load_store_type): Update calls to get_group_load_store_type
22508 and vect_use_strided_gather_scatters_p.
22510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22511 Alan Hayward <alan.hayward@arm.com>
22512 David Sherwood <david.sherwood@arm.com>
22514 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22515 optional tree argument.
22516 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22518 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22519 but continue to use the current value as a fallback.
22520 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22521 to compare the updates.
22522 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22523 (get_load_store_type): Use it when handling a strided access.
22524 (vect_get_strided_load_store_ops): New function.
22525 (vect_get_data_ptr_increment): Likewise.
22526 (vectorizable_load): Handle strided gather loads. Always pass
22527 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22530 Alan Hayward <alan.hayward@arm.com>
22531 David Sherwood <david.sherwood@arm.com>
22533 * doc/md.texi (gather_load@var{m}): Document.
22534 (mask_gather_load@var{m}): Likewise.
22535 * genopinit.c (main): Add supports_vec_gather_load and
22536 supports_vec_gather_load_cached to target_optabs.
22537 * optabs-tree.c (init_tree_optimization_optabs): Use
22538 ggc_cleared_alloc to allocate target_optabs.
22539 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22540 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22542 * internal-fn.h (internal_load_fn_p): Declare.
22543 (internal_gather_scatter_fn_p): Likewise.
22544 (internal_fn_mask_index): Likewise.
22545 (internal_gather_scatter_fn_supported_p): Likewise.
22546 * internal-fn.c (gather_load_direct): New macro.
22547 (expand_gather_load_optab_fn): New function.
22548 (direct_gather_load_optab_supported_p): New macro.
22549 (direct_internal_fn_optab): New function.
22550 (internal_load_fn_p): Likewise.
22551 (internal_gather_scatter_fn_p): Likewise.
22552 (internal_fn_mask_index): Likewise.
22553 (internal_gather_scatter_fn_supported_p): Likewise.
22554 * optabs-query.c (supports_at_least_one_mode_p): New function.
22555 (supports_vec_gather_load_p): Likewise.
22556 * optabs-query.h (supports_vec_gather_load_p): Declare.
22557 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22558 and memory_type field.
22559 (NUM_PATTERNS): Bump to 15.
22560 * tree-vect-data-refs.c: Include internal-fn.h.
22561 (vect_gather_scatter_fn_p): New function.
22562 (vect_describe_gather_scatter_call): Likewise.
22563 (vect_check_gather_scatter): Try using internal functions for
22564 gather loads. Recognize existing calls to a gather load function.
22565 (vect_analyze_data_refs): Consider using gather loads if
22566 supports_vec_gather_load_p.
22567 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22568 (vect_get_gather_scatter_offset_type): Likewise.
22569 (vect_convert_mask_for_vectype): Likewise.
22570 (vect_add_conversion_to_patterm): Likewise.
22571 (vect_try_gather_scatter_pattern): Likewise.
22572 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22573 (vect_vect_recog_func_ptrs): Add it.
22574 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22575 internal_fn_mask_index and internal_gather_scatter_fn_p.
22576 (check_load_store_masking): Take the gather_scatter_info as an
22577 argument and handle gather loads.
22578 (vect_get_gather_scatter_ops): New function.
22579 (vectorizable_call): Check internal_load_fn_p.
22580 (vectorizable_load): Likewise. Handle gather load internal
22582 (vectorizable_store): Update call to check_load_store_masking.
22583 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22584 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22585 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22586 (aarch64_gather_scale_operand_d): New predicates.
22587 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22588 (mask_gather_load<mode>): New insns.
22590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22591 Alan Hayward <alan.hayward@arm.com>
22592 David Sherwood <david.sherwood@arm.com>
22594 * optabs.def (fold_left_plus_optab): New optab.
22595 * doc/md.texi (fold_left_plus_@var{m}): Document.
22596 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22597 * internal-fn.c (fold_left_direct): Define.
22598 (expand_fold_left_optab_fn): Likewise.
22599 (direct_fold_left_optab_supported_p): Likewise.
22600 * fold-const-call.c (fold_const_fold_left): New function.
22601 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22602 * tree-parloops.c (valid_reduction_p): New function.
22603 (gather_scalar_reductions): Use it.
22604 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22605 (vect_finish_replace_stmt): Declare.
22606 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22607 (needs_fold_left_reduction_p): New function, split out from...
22608 (vect_is_simple_reduction): ...here. Accept reductions that
22609 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22610 (vect_force_simple_reduction): Also store the reduction type in
22611 the assignment's STMT_VINFO_REDUC_TYPE.
22612 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22613 (merge_with_identity): New function.
22614 (vect_expand_fold_left): Likewise.
22615 (vectorize_fold_left_reduction): Likewise.
22616 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22617 scalar phi in place for it. Check for target support and reject
22618 cases that would reassociate the operation. Defer the transform
22619 phase to vectorize_fold_left_reduction.
22620 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22621 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22622 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22626 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22627 call to ifc_temp_var.
22629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22630 Alan Hayward <alan.hayward@arm.com>
22631 David Sherwood <david.sherwood@arm.com>
22633 * target.def (legitimize_address_displacement): Take the original
22634 offset as a poly_int.
22635 * targhooks.h (default_legitimize_address_displacement): Update
22637 * targhooks.c (default_legitimize_address_displacement): Likewise.
22638 * doc/tm.texi: Regenerate.
22639 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22640 as an argument, moving assert of ad->disp == ad->disp_term to...
22641 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22642 Try calling targetm.legitimize_address_displacement before expanding
22643 the address rather than afterwards, and adjust for the new interface.
22644 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22645 Match the new hook interface. Handle SVE addresses.
22646 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22647 new hook interface.
22649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22651 * Makefile.in (OBJS): Add early-remat.o.
22652 * target.def (select_early_remat_modes): New hook.
22653 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22654 * doc/tm.texi: Regenerate.
22655 * targhooks.h (default_select_early_remat_modes): Declare.
22656 * targhooks.c (default_select_early_remat_modes): New function.
22657 * timevar.def (TV_EARLY_REMAT): New timevar.
22658 * passes.def (pass_early_remat): New pass.
22659 * tree-pass.h (make_pass_early_remat): Declare.
22660 * early-remat.c: New file.
22661 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22663 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22666 Alan Hayward <alan.hayward@arm.com>
22667 David Sherwood <david.sherwood@arm.com>
22669 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22670 vfm1 with a bound_epilog parameter.
22671 (vect_do_peeling): Update calls accordingly, and move the prologue
22672 call earlier in the function. Treat the base bound_epilog as 0 for
22673 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22674 this base when peeling for gaps.
22675 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22676 with fully-masked loops.
22677 (vect_estimate_min_profitable_iters): Handle the single peeled
22678 iteration in that case.
22680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22681 Alan Hayward <alan.hayward@arm.com>
22682 David Sherwood <david.sherwood@arm.com>
22684 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22685 single-element interleaving even if the size is not a power of 2.
22686 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22687 accesses for single-element interleaving if the group size is
22690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22691 Alan Hayward <alan.hayward@arm.com>
22692 David Sherwood <david.sherwood@arm.com>
22694 * doc/md.texi (fold_extract_last_@var{m}): Document.
22695 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22696 * optabs.def (fold_extract_last_optab): New optab.
22697 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22698 * internal-fn.c (fold_extract_direct): New macro.
22699 (expand_fold_extract_optab_fn): Likewise.
22700 (direct_fold_extract_optab_supported_p): Likewise.
22701 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22702 * tree-vect-loop.c (vect_model_reduction_cost): Handle
22703 EXTRACT_LAST_REDUCTION.
22704 (get_initial_def_for_reduction): Do not create an initial vector
22705 for EXTRACT_LAST_REDUCTION reductions.
22706 (vectorizable_reduction): Leave the scalar phi in place for
22707 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
22708 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
22709 epilogue code for EXTRACT_LAST_REDUCTION and defer the
22710 transform phase to vectorizable_condition.
22711 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22713 (vect_finish_stmt_generation): ...here.
22714 (vect_finish_replace_stmt): New function.
22715 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22716 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22718 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22721 Alan Hayward <alan.hayward@arm.com>
22722 David Sherwood <david.sherwood@arm.com>
22724 * doc/md.texi (extract_last_@var{m}): Document.
22725 * optabs.def (extract_last_optab): New optab.
22726 * internal-fn.def (EXTRACT_LAST): New internal function.
22727 * internal-fn.c (cond_unary_direct): New macro.
22728 (expand_cond_unary_optab_fn): Likewise.
22729 (direct_cond_unary_optab_supported_p): Likewise.
22730 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22731 loops using EXTRACT_LAST.
22732 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22733 (extract_last_<mode>): ...this optab.
22734 (vec_extract<mode><Vel>): Update accordingly.
22736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22737 Alan Hayward <alan.hayward@arm.com>
22738 David Sherwood <david.sherwood@arm.com>
22740 * target.def (empty_mask_is_expensive): New hook.
22741 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22742 * doc/tm.texi: Regenerate.
22743 * targhooks.h (default_empty_mask_is_expensive): Declare.
22744 * targhooks.c (default_empty_mask_is_expensive): New function.
22745 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22746 if the target says that empty masks are expensive.
22747 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22749 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22752 Alan Hayward <alan.hayward@arm.com>
22753 David Sherwood <david.sherwood@arm.com>
22755 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22756 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22757 (vect_use_loop_mask_for_alignment_p): New function.
22758 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22759 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22760 niters_skip argument. Make sure that the first niters_skip elements
22761 of the first iteration are inactive.
22762 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22763 Update call to vect_set_loop_masks_directly.
22764 (get_misalign_in_elems): New function, split out from...
22765 (vect_gen_prolog_loop_niters): ...here.
22766 (vect_update_init_of_dr): Take a code argument that specifies whether
22767 the adjustment should be added or subtracted.
22768 (vect_update_init_of_drs): Likewise.
22769 (vect_prepare_for_masked_peels): New function.
22770 (vect_do_peeling): Skip prologue peeling if we're using a mask
22771 instead. Update call to vect_update_inits_of_drs.
22772 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22774 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22775 alignment. Do not include the number of peeled iterations in
22776 the minimum threshold in that case.
22777 (vectorizable_induction): Adjust the start value down by
22778 LOOP_VINFO_MASK_SKIP_NITERS iterations.
22779 (vect_transform_loop): Call vect_prepare_for_masked_peels.
22780 Take the number of skipped iterations into account when calculating
22782 * tree-vect-stmts.c (vect_gen_while_not): New function.
22784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22785 Alan Hayward <alan.hayward@arm.com>
22786 David Sherwood <david.sherwood@arm.com>
22788 * doc/sourcebuild.texi (vect_fully_masked): Document.
22789 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22790 default value to 0.
22791 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22793 (vect_analyze_loop_2): ...here. Don't check the vectorization
22794 factor against the number of loop iterations if the loop is
22797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22798 Alan Hayward <alan.hayward@arm.com>
22799 David Sherwood <david.sherwood@arm.com>
22801 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22802 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22803 (dump_groups): Update accordingly.
22804 (iv_use::mem_type): New member variable.
22805 (address_p): New function.
22806 (record_use): Add a mem_type argument and initialize the new
22808 (record_group_use): Add a mem_type argument. Use address_p.
22809 Remove obsolete null checks of base_object. Update call to record_use.
22810 (find_interesting_uses_op): Update call to record_group_use.
22811 (find_interesting_uses_cond): Likewise.
22812 (find_interesting_uses_address): Likewise.
22813 (get_mem_type_for_internal_fn): New function.
22814 (find_address_like_use): Likewise.
22815 (find_interesting_uses_stmt): Try find_address_like_use before
22816 calling find_interesting_uses_op.
22817 (addr_offset_valid_p): Use the iv mem_type field as the type
22818 of the addressed memory.
22819 (add_autoinc_candidates): Likewise.
22820 (get_address_cost): Likewise.
22821 (split_small_address_groups_p): Use address_p.
22822 (split_address_groups): Likewise.
22823 (add_iv_candidate_for_use): Likewise.
22824 (autoinc_possible_for_pair): Likewise.
22825 (rewrite_groups): Likewise.
22826 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22827 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22828 (get_alias_ptr_type_for_ptr_address): New function.
22829 (rewrite_use_address): Rewrite address uses in calls that were
22830 identified by find_address_like_use.
22832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22833 Alan Hayward <alan.hayward@arm.com>
22834 David Sherwood <david.sherwood@arm.com>
22836 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22838 * gimple-expr.h (is_gimple_addressable: Likewise.
22839 * gimple-expr.c (is_gimple_address): Likewise.
22840 * internal-fn.c (expand_call_mem_ref): New function.
22841 (expand_mask_load_optab_fn): Use it.
22842 (expand_mask_store_optab_fn): Likewise.
22844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22845 Alan Hayward <alan.hayward@arm.com>
22846 David Sherwood <david.sherwood@arm.com>
22848 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22849 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22850 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22851 (cond_umax@var{mode}): Document.
22852 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22853 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22854 (cond_umin_optab, cond_umax_optab): New optabs.
22855 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22856 (COND_IOR, COND_XOR): New internal functions.
22857 * internal-fn.h (get_conditional_internal_fn): Declare.
22858 * internal-fn.c (cond_binary_direct): New macro.
22859 (expand_cond_binary_optab_fn): Likewise.
22860 (direct_cond_binary_optab_supported_p): Likewise.
22861 (get_conditional_internal_fn): New function.
22862 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22863 Cope with reduction statements that are vectorized as calls rather
22865 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22866 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22867 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22868 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22869 (UNSPEC_COND_EOR): New unspecs.
22870 (optab): Add mappings for them.
22871 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22872 (sve_int_op, sve_fp_op): New int attributes.
22874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22875 Alan Hayward <alan.hayward@arm.com>
22876 David Sherwood <david.sherwood@arm.com>
22878 * optabs.def (while_ult_optab): New optab.
22879 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22880 * internal-fn.def (WHILE_ULT): New internal function.
22881 * internal-fn.h (direct_internal_fn_supported_p): New override
22882 that takes two types as argument.
22883 * internal-fn.c (while_direct): New macro.
22884 (expand_while_optab_fn): New function.
22885 (convert_optab_supported_p): Likewise.
22886 (direct_while_optab_supported_p): New macro.
22887 * wide-int.h (wi::udiv_ceil): New function.
22888 * tree-vectorizer.h (rgroup_masks): New structure.
22889 (vec_loop_masks): New typedef.
22890 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22891 and fully_masked_p.
22892 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22893 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22894 (vect_max_vf): New function.
22895 (slpeel_make_loop_iterate_ntimes): Delete.
22896 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22897 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22898 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22899 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22900 internal-fn.h, stor-layout.h and optabs-query.h.
22901 (vect_set_loop_mask): New function.
22902 (add_preheader_seq): Likewise.
22903 (add_header_seq): Likewise.
22904 (interleave_supported_p): Likewise.
22905 (vect_maybe_permute_loop_masks): Likewise.
22906 (vect_set_loop_masks_directly): Likewise.
22907 (vect_set_loop_condition_masked): Likewise.
22908 (vect_set_loop_condition_unmasked): New function, split out from
22909 slpeel_make_loop_iterate_ntimes.
22910 (slpeel_make_loop_iterate_ntimes): Rename to..
22911 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22912 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22913 (vect_do_peeling): Update call accordingly.
22914 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22916 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22917 mask_compare_type, can_fully_mask_p and fully_masked_p.
22918 (release_vec_loop_masks): New function.
22919 (_loop_vec_info): Use it to free the loop masks.
22920 (can_produce_all_loop_masks_p): New function.
22921 (vect_get_max_nscalars_per_iter): Likewise.
22922 (vect_verify_full_masking): Likewise.
22923 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22924 retries, and free the mask rgroups before retrying. Check loop-wide
22925 reasons for disallowing fully-masked loops. Make the final decision
22926 about whether use a fully-masked loop or not.
22927 (vect_estimate_min_profitable_iters): Do not assume that peeling
22928 for the number of iterations will be needed for fully-masked loops.
22929 (vectorizable_reduction): Disable fully-masked loops.
22930 (vectorizable_live_operation): Likewise.
22931 (vect_halve_mask_nunits): New function.
22932 (vect_double_mask_nunits): Likewise.
22933 (vect_record_loop_mask): Likewise.
22934 (vect_get_loop_mask): Likewise.
22935 (vect_transform_loop): Handle the case in which the final loop
22936 iteration might handle a partial vector. Call vect_set_loop_condition
22937 instead of slpeel_make_loop_iterate_ntimes.
22938 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22939 (check_load_store_masking): New function.
22940 (prepare_load_store_mask): Likewise.
22941 (vectorizable_store): Handle fully-masked loops.
22942 (vectorizable_load): Likewise.
22943 (supportable_widening_operation): Use vect_halve_mask_nunits for
22945 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22946 (vect_gen_while): New function.
22947 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22948 (aarch64_uqdec<mode>): New insn.
22950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22951 Alan Hayward <alan.hayward@arm.com>
22952 David Sherwood <david.sherwood@arm.com>
22954 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22955 (reduc_xor_scal_optab): New optabs.
22956 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22957 (reduc_xor_scal_@var{m}): Document.
22958 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22959 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22960 internal functions.
22961 * fold-const-call.c (fold_const_call): Handle them.
22962 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22963 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22964 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22965 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22966 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22967 (UNSPEC_XORV): New unspecs.
22968 (optab): Add entries for them.
22969 (BITWISEV): New int iterator.
22970 (bit_reduc_op): New int attributes.
22972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22973 Alan Hayward <alan.hayward@arm.com>
22974 David Sherwood <david.sherwood@arm.com>
22976 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22977 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22978 * optabs.def (vec_shl_insert_optab): New optab.
22979 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22980 (duplicate_and_interleave): Likewise.
22981 * tree-vect-loop.c: Include internal-fn.h.
22982 (neutral_op_for_slp_reduction): New function, split out from
22983 get_initial_defs_for_reduction.
22984 (get_initial_def_for_reduction): Handle option 2 for variable-length
22985 vectors by loading the neutral value into a vector and then shifting
22986 the initial value into element 0.
22987 (get_initial_defs_for_reduction): Replace the code argument with
22988 the neutral value calculated by neutral_op_for_slp_reduction.
22989 Use gimple_build_vector for constant-length vectors.
22990 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22991 but the first group_size elements have a neutral value.
22992 Use duplicate_and_interleave otherwise.
22993 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22994 Update call to get_initial_defs_for_reduction. Handle SLP
22995 reductions for variable-length vectors by creating one vector
22996 result for each scalar result, with the elements associated
22997 with other scalar results stubbed out with the neutral value.
22998 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22999 Require IFN_VEC_SHL_INSERT for double reductions on
23000 variable-length vectors, or SLP reductions that have
23001 a neutral value. Require can_duplicate_and_interleave_p
23002 support for variable-length unchained SLP reductions if there
23003 is no neutral value, such as for MIN/MAX reductions. Also require
23004 the number of vector elements to be a multiple of the number of
23005 SLP statements when doing variable-length unchained SLP reductions.
23006 Update call to vect_create_epilog_for_reduction.
23007 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
23008 and remove initial values.
23009 (duplicate_and_interleave): Make public.
23010 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
23011 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
23013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23014 Alan Hayward <alan.hayward@arm.com>
23015 David Sherwood <david.sherwood@arm.com>
23017 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
23018 (can_duplicate_and_interleave_p): New function.
23019 (vect_get_and_check_slp_defs): Take the vector of statements
23020 rather than just the current one. Remove excess parentheses.
23021 Restriction rejectinon of vect_constant_def and vect_external_def
23022 for variable-length vectors to boolean types, or types for which
23023 can_duplicate_and_interleave_p is false.
23024 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
23025 (duplicate_and_interleave): New function.
23026 (vect_get_constant_vectors): Use gimple_build_vector for
23027 constant-length vectors and suitable variable-length constant
23028 vectors. Use duplicate_and_interleave for other variable-length
23029 vectors. Don't defer the update when inserting new statements.
23031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23032 Alan Hayward <alan.hayward@arm.com>
23033 David Sherwood <david.sherwood@arm.com>
23035 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
23036 min_profitable_iters doesn't go negative.
23038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23039 Alan Hayward <alan.hayward@arm.com>
23040 David Sherwood <david.sherwood@arm.com>
23042 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
23043 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
23044 * optabs.def (vec_mask_load_lanes_optab): New optab.
23045 (vec_mask_store_lanes_optab): Likewise.
23046 * internal-fn.def (MASK_LOAD_LANES): New internal function.
23047 (MASK_STORE_LANES): Likewise.
23048 * internal-fn.c (mask_load_lanes_direct): New macro.
23049 (mask_store_lanes_direct): Likewise.
23050 (expand_mask_load_optab_fn): Handle masked operations.
23051 (expand_mask_load_lanes_optab_fn): New macro.
23052 (expand_mask_store_optab_fn): Handle masked operations.
23053 (expand_mask_store_lanes_optab_fn): New macro.
23054 (direct_mask_load_lanes_optab_supported_p): Likewise.
23055 (direct_mask_store_lanes_optab_supported_p): Likewise.
23056 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
23058 (vect_load_lanes_supported): Likewise.
23059 * tree-vect-data-refs.c (strip_conversion): New function.
23060 (can_group_stmts_p): Likewise.
23061 (vect_analyze_data_ref_accesses): Use it instead of checking
23062 for a pair of assignments.
23063 (vect_store_lanes_supported): Take a masked_p parameter.
23064 (vect_load_lanes_supported): Likewise.
23065 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
23066 vect_store_lanes_supported and vect_load_lanes_supported.
23067 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
23068 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
23069 parameter. Don't allow gaps for masked accesses.
23070 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
23071 and vect_load_lanes_supported.
23072 (get_load_store_type): Take a masked_p parameter and update
23073 call to get_group_load_store_type.
23074 (vectorizable_store): Update call to get_load_store_type.
23075 Handle IFN_MASK_STORE_LANES.
23076 (vectorizable_load): Update call to get_load_store_type.
23077 Handle IFN_MASK_LOAD_LANES.
23079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23080 Alan Hayward <alan.hayward@arm.com>
23081 David Sherwood <david.sherwood@arm.com>
23083 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
23085 * config/aarch64/aarch64-protos.h
23086 (aarch64_sve_struct_memory_operand_p): Declare.
23087 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
23088 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
23089 (VPRED, vpred): Handle SVE structure modes.
23090 * config/aarch64/constraints.md (Utx): New constraint.
23091 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
23092 (aarch64_sve_struct_nonimmediate_operand): New predicates.
23093 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
23094 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
23095 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
23096 structure modes. Split into pieces after RA.
23097 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
23098 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
23100 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
23101 SVE structure modes.
23102 (aarch64_classify_address): Likewise.
23103 (sizetochar): Move earlier in file.
23104 (aarch64_print_operand): Handle SVE register lists.
23105 (aarch64_array_mode): New function.
23106 (aarch64_sve_struct_memory_operand_p): Likewise.
23107 (TARGET_ARRAY_MODE): Redefine.
23109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23110 Alan Hayward <alan.hayward@arm.com>
23111 David Sherwood <david.sherwood@arm.com>
23113 * target.def (array_mode): New target hook.
23114 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
23115 * doc/tm.texi: Regenerate.
23116 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
23117 * hooks.c (hook_optmode_mode_uhwi_none): New function.
23118 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
23119 targetm.array_mode.
23120 * stor-layout.c (mode_for_array): Likewise. Support polynomial
23123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23124 Alan Hayward <alan.hayward@arm.com>
23125 David Sherwood <david.sherwood@arm.com>
23127 * fold-const.c (fold_binary_loc): Check the argument types
23128 rather than the result type when testing for a vector operation.
23130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23132 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
23133 * doc/tm.texi: Regenerate.
23135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23136 Alan Hayward <alan.hayward@arm.com>
23137 David Sherwood <david.sherwood@arm.com>
23139 * doc/invoke.texi (-msve-vector-bits=): Document new option.
23140 (sve): Document new AArch64 extension.
23141 * doc/md.texi (w): Extend the description of the AArch64
23142 constraint to include SVE vectors.
23143 (Upl, Upa): Document new AArch64 predicate constraints.
23144 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
23146 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
23147 (msve-vector-bits=): New option.
23148 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
23149 SVE when these are disabled.
23150 (sve): New extension.
23151 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
23152 modes. Adjust their number of units based on aarch64_sve_vg.
23153 (MAX_BITSIZE_MODE_ANY_MODE): Define.
23154 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
23155 aarch64_addr_query_type.
23156 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
23157 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
23158 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
23159 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
23160 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
23161 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
23162 (aarch64_simd_imm_zero_p): Delete.
23163 (aarch64_check_zero_based_sve_index_immediate): Declare.
23164 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23165 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23166 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23167 (aarch64_sve_float_mul_immediate_p): Likewise.
23168 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23169 rather than an rtx.
23170 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
23171 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
23172 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
23173 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
23174 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
23175 (aarch64_regmode_natural_size): Likewise.
23176 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
23177 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
23179 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
23180 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
23181 for VG and the SVE predicate registers.
23182 (V_ALIASES): Add a "z"-prefixed alias.
23183 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
23184 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
23185 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
23186 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
23187 (REG_CLASS_NAMES): Add entries for them.
23188 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
23189 and the predicate registers.
23190 (aarch64_sve_vg): Declare.
23191 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
23192 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
23193 (REGMODE_NATURAL_SIZE): Define.
23194 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
23196 * config/aarch64/aarch64.c: Include cfgrtl.h.
23197 (simd_immediate_info): Add a constructor for series vectors,
23198 and an associated step field.
23199 (aarch64_sve_vg): New variable.
23200 (aarch64_dbx_register_number): Handle VG and the predicate registers.
23201 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
23202 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
23203 (VEC_ANY_DATA, VEC_STRUCT): New constants.
23204 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
23205 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
23206 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
23207 (aarch64_get_mask_mode): New functions.
23208 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
23209 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
23210 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
23211 predicate modes and predicate registers. Explicitly restrict
23212 GPRs to modes of 16 bytes or smaller. Only allow FP registers
23213 to store a vector mode if it is recognized by
23214 aarch64_classify_vector_mode.
23215 (aarch64_regmode_natural_size): New function.
23216 (aarch64_hard_regno_caller_save_mode): Return the original mode
23218 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
23219 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
23220 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
23221 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
23223 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
23224 does not overlap dest if the function is frame-related. Handle
23226 (aarch64_split_add_offset): New function.
23227 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
23228 them aarch64_add_offset.
23229 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
23230 and update call to aarch64_sub_sp.
23231 (aarch64_add_cfa_expression): New function.
23232 (aarch64_expand_prologue): Pass extra temporary registers to the
23233 functions above. Handle the case in which we need to emit new
23234 DW_CFA_expressions for registers that were originally saved
23235 relative to the stack pointer, but now have to be expressed
23236 relative to the frame pointer.
23237 (aarch64_output_mi_thunk): Pass extra temporary registers to the
23239 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
23240 IP0 and IP1 values for SVE frames.
23241 (aarch64_expand_vec_series): New function.
23242 (aarch64_expand_sve_widened_duplicate): Likewise.
23243 (aarch64_expand_sve_const_vector): Likewise.
23244 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
23245 Handle SVE constants. Use emit_move_insn to move a force_const_mem
23246 into the register, rather than emitting a SET directly.
23247 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
23248 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
23249 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
23250 (offset_9bit_signed_scaled_p): New functions.
23251 (aarch64_replicate_bitmask_imm): New function.
23252 (aarch64_bitmask_imm): Use it.
23253 (aarch64_cannot_force_const_mem): Reject expressions involving
23254 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
23255 (aarch64_classify_index): Handle SVE indices, by requiring
23256 a plain register index with a scale that matches the element size.
23257 (aarch64_classify_address): Handle SVE addresses. Assert that
23258 the mode of the address is VOIDmode or an integer mode.
23259 Update call to aarch64_classify_symbol.
23260 (aarch64_classify_symbolic_expression): Update call to
23261 aarch64_classify_symbol.
23262 (aarch64_const_vec_all_in_range_p): New function.
23263 (aarch64_print_vector_float_operand): Likewise.
23264 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
23265 "vN" for FP registers with SVE modes. Handle (const ...) vectors
23266 and the FP immediates 1.0 and 0.5.
23267 (aarch64_print_address_internal): Handle SVE addresses.
23268 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
23269 (aarch64_regno_regclass): Handle predicate registers.
23270 (aarch64_secondary_reload): Handle big-endian reloads of SVE
23272 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
23273 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
23274 (aarch64_convert_sve_vector_bits): New function.
23275 (aarch64_override_options): Use it to handle -msve-vector-bits=.
23276 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23277 rather than an rtx.
23278 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
23279 Handle SVE vector and predicate modes. Accept VL-based constants
23280 that need only one temporary register, and VL offsets that require
23281 no temporary registers.
23282 (aarch64_conditional_register_usage): Mark the predicate registers
23283 as fixed if SVE isn't available.
23284 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
23285 Return true for SVE vector and predicate modes.
23286 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
23287 rather than an unsigned int. Handle SVE modes.
23288 (aarch64_preferred_simd_mode): Update call accordingly. Handle
23290 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
23292 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23293 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23294 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23295 (aarch64_sve_float_mul_immediate_p): New functions.
23296 (aarch64_sve_valid_immediate): New function.
23297 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
23298 Explicitly reject structure modes. Check for INDEX constants.
23299 Handle PTRUE and PFALSE constants.
23300 (aarch64_check_zero_based_sve_index_immediate): New function.
23301 (aarch64_simd_imm_zero_p): Delete.
23302 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
23303 vector modes. Accept constants in the range of CNT[BHWD].
23304 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
23305 ask for an Advanced SIMD mode.
23306 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
23307 (aarch64_simd_vector_alignment): Handle SVE predicates.
23308 (aarch64_vectorize_preferred_vector_alignment): New function.
23309 (aarch64_simd_vector_alignment_reachable): Use it instead of
23311 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
23312 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
23314 (MAX_VECT_LEN): Delete.
23315 (expand_vec_perm_d): Add a vec_flags field.
23316 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
23317 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
23318 (aarch64_evpc_ext): Don't apply a big-endian lane correction
23320 (aarch64_evpc_rev): Rename to...
23321 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
23322 (aarch64_evpc_rev_global): New function.
23323 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
23324 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
23326 (aarch64_evpc_sve_tbl): New function.
23327 (aarch64_expand_vec_perm_const_1): Update after rename of
23328 aarch64_evpc_rev. Handle SVE permutes too, trying
23329 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
23330 than aarch64_evpc_tbl.
23331 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
23332 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
23333 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
23334 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
23335 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
23336 (aarch64_expand_sve_vcond): New functions.
23337 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
23338 of aarch64_vector_mode_p.
23339 (aarch64_dwarf_poly_indeterminate_value): New function.
23340 (aarch64_compute_pressure_classes): Likewise.
23341 (aarch64_can_change_mode_class): Likewise.
23342 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
23343 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
23344 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
23345 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
23346 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
23347 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
23348 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
23349 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
23351 (Dn, Dl, Dr): Accept const as well as const_vector.
23352 (Dz): Likewise. Compare against CONST0_RTX.
23353 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
23354 of "vector" where appropriate.
23355 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
23356 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
23357 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
23358 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
23359 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
23360 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
23361 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
23362 (v_int_equiv): Extend to SVE modes.
23363 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
23365 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
23366 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
23367 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
23368 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
23369 (SVE_COND_FP_CMP): New int iterators.
23370 (perm_hilo): Handle the new unpack unspecs.
23371 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
23373 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
23374 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
23375 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
23376 (aarch64_equality_operator, aarch64_constant_vector_operand)
23377 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
23378 (aarch64_sve_nonimmediate_operand): Likewise.
23379 (aarch64_sve_general_operand): Likewise.
23380 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
23381 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
23382 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
23383 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
23384 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
23385 (aarch64_sve_float_arith_immediate): Likewise.
23386 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
23387 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
23388 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
23389 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
23390 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
23391 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
23392 (aarch64_sve_float_arith_operand): Likewise.
23393 (aarch64_sve_float_arith_with_sub_operand): Likewise.
23394 (aarch64_sve_float_mul_operand): Likewise.
23395 (aarch64_sve_vec_perm_operand): Likewise.
23396 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
23397 (aarch64_mov_operand): Accept const_poly_int and const_vector.
23398 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
23399 as well as const_vector.
23400 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
23401 in file. Use CONST0_RTX and CONSTM1_RTX.
23402 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
23403 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
23404 Use aarch64_simd_imm_zero.
23405 * config/aarch64/aarch64-sve.md: New file.
23406 * config/aarch64/aarch64.md: Include it.
23407 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
23408 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
23409 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
23410 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
23411 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
23412 (sve): New attribute.
23413 (enabled): Disable instructions with the sve attribute unless
23415 (movqi, movhi): Pass CONST_POLY_INT operaneds through
23416 aarch64_expand_mov_immediate.
23417 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
23418 CNT[BHSD] immediates.
23419 (movti): Split CONST_POLY_INT moves into two halves.
23420 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
23421 Split additions that need a temporary here if the destination
23422 is the stack pointer.
23423 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
23424 (*add<mode>3_poly_1): New instruction.
23425 (set_clobber_cc): New expander.
23427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23429 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
23430 parameter and use it instead of GET_MODE_SIZE (innermode). Use
23431 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
23432 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
23433 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
23434 Change innermode from fixed_mode_size to machine_mode.
23435 (simplify_subreg): Update call accordingly. Handle a constant-sized
23436 subreg of a variable-length CONST_VECTOR.
23438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23439 Alan Hayward <alan.hayward@arm.com>
23440 David Sherwood <david.sherwood@arm.com>
23442 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23443 (add_offset_to_base): New function, split out from...
23444 (create_mem_ref): ...here. When handling a scale other than 1,
23445 check first whether the address is valid without the offset.
23446 Add it into the base if so, leaving the index and scale as-is.
23448 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23451 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23452 fold_for_warn before checking if arg2 is INTEGER_CST.
23454 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
23456 * config/rs6000/predicates.md (load_multiple_operation): Delete.
23457 (store_multiple_operation): Delete.
23458 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23459 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23460 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23461 guarded by TARGET_STRING.
23462 (rs6000_output_load_multiple): Delete.
23463 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23464 OPTION_MASK_STRING / TARGET_STRING handling.
23465 (print_operand) <'N', 'O'>: Add comment that these are unused now.
23466 (const rs6000_opt_masks) <"string">: Change mask to 0.
23467 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23468 (MASK_STRING): Delete.
23469 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23471 (load_multiple): Delete.
23478 (store_multiple): Delete.
23485 (movmemsi_8reg): Delete.
23486 (corresponding unnamed define_insn): Delete.
23487 (movmemsi_6reg): Delete.
23488 (corresponding unnamed define_insn): Delete.
23489 (movmemsi_4reg): Delete.
23490 (corresponding unnamed define_insn): Delete.
23491 (movmemsi_2reg): Delete.
23492 (corresponding unnamed define_insn): Delete.
23493 (movmemsi_1reg): Delete.
23494 (corresponding unnamed define_insn): Delete.
23495 * config/rs6000/rs6000.opt (mno-string): New.
23496 (mstring): Replace by deprecation warning stub.
23497 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23499 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23501 * regrename.c (regrename_do_replace): If replacing the same
23502 reg multiple times, try to reuse last created gen_raw_REG.
23505 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23506 main to workaround a bug in GDB.
23508 2018-01-12 Tom de Vries <tom@codesourcery.com>
23511 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23513 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23515 PR rtl-optimization/80481
23516 * ira-color.c (get_cap_member): New function.
23517 (allocnos_conflict_by_live_ranges_p): Use it.
23518 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23519 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23521 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23524 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23525 (*saddl_se_1): Ditto.
23526 (*ssubsi_1): Ditto.
23527 (*ssubl_se_1): Ditto.
23529 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23531 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23532 rather than wi::to_widest for DR_INITs.
23533 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23534 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23535 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23537 (vect_analyze_group_access_1): Note that here.
23539 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23541 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23542 polynomial type sizes.
23544 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23546 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23547 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23548 (gimple_add_tmp_var): Likewise.
23550 2018-01-12 Martin Liska <mliska@suse.cz>
23552 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23553 (gimple_alloc_sizes): Likewise.
23554 (dump_gimple_statistics): Use PRIu64 in printf format.
23555 * gimple.h: Change uint64_t to int.
23557 2018-01-12 Martin Liska <mliska@suse.cz>
23559 * tree-core.h: Use uint64_t instead of int.
23560 * tree.c (tree_node_counts): Likewise.
23561 (tree_node_sizes): Likewise.
23562 (dump_tree_statistics): Use PRIu64 in printf format.
23564 2018-01-12 Martin Liska <mliska@suse.cz>
23566 * Makefile.in: As qsort_chk is implemented in vec.c, add
23567 vec.o to linkage of gencfn-macros.
23568 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23569 passing the info to record_node_allocation_statistics.
23570 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23572 * ggc-common.c (struct ggc_usage): Add operator== and use
23573 it in operator< and compare function.
23574 * mem-stats.h (struct mem_usage): Likewise.
23575 * vec.c (struct vec_usage): Remove operator< and compare
23576 function. Can be simply inherited.
23578 2018-01-12 Martin Jambor <mjambor@suse.cz>
23581 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23582 * tree-ssa-math-opts.c: Include domwalk.h.
23583 (convert_mult_to_fma_1): New function.
23584 (fma_transformation_info): New type.
23585 (fma_deferring_state): Likewise.
23586 (cancel_fma_deferring): New function.
23587 (result_of_phi): Likewise.
23588 (last_fma_candidate_feeds_initial_phi): Likewise.
23589 (convert_mult_to_fma): Added deferring logic, split actual
23590 transformation to convert_mult_to_fma_1.
23591 (math_opts_dom_walker): New type.
23592 (math_opts_dom_walker::after_dom_children): New method, body moved
23593 here from pass_optimize_widening_mul::execute, added deferring logic
23595 (pass_optimize_widening_mul::execute): Moved most of code to
23596 math_opts_dom_walker::after_dom_children.
23597 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23598 * config/i386/i386.c (ix86_option_override_internal): Added
23599 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23601 2018-01-12 Richard Biener <rguenther@suse.de>
23604 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23605 inline instance vars.
23607 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23610 * config/rx/rx.c (rx_is_restricted_memory_address):
23611 Handle SUBREG case.
23613 2018-01-12 Richard Biener <rguenther@suse.de>
23615 PR tree-optimization/80846
23616 * target.def (split_reduction): New target hook.
23617 * targhooks.c (default_split_reduction): New function.
23618 * targhooks.h (default_split_reduction): Declare.
23619 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23620 target requests first reduce vectors by combining low and high
23622 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23623 (get_vectype_for_scalar_type_and_size): Export.
23624 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23625 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23626 * doc/tm.texi: Regenerate.
23627 * config/i386/i386.c (ix86_split_reduction): Implement
23628 TARGET_VECTORIZE_SPLIT_REDUCTION.
23630 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23633 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23634 in PIC mode except for TARGET_VXWORKS_RTP.
23635 * config/sparc/sparc.c: Include cfgrtl.h.
23636 (TARGET_INIT_PIC_REG): Define.
23637 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23638 (sparc_pic_register_p): New predicate.
23639 (sparc_legitimate_address_p): Use it.
23640 (sparc_legitimize_pic_address): Likewise.
23641 (sparc_delegitimize_address): Likewise.
23642 (sparc_mode_dependent_address_p): Likewise.
23643 (gen_load_pcrel_sym): Remove 4th parameter.
23644 (load_got_register): Adjust call to above. Remove obsolete stuff.
23645 (sparc_expand_prologue): Do not call load_got_register here.
23646 (sparc_flat_expand_prologue): Likewise.
23647 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23648 (sparc_use_pseudo_pic_reg): New function.
23649 (sparc_init_pic_reg): Likewise.
23650 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23651 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23653 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23655 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23656 Add item for branch_cost.
23658 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23660 PR rtl-optimization/83565
23661 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23662 not extend the result to a larger mode for rotate operations.
23663 (num_sign_bit_copies1): Likewise.
23665 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23668 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23670 Use values-Xc.o for -pedantic.
23671 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23673 2018-01-12 Martin Liska <mliska@suse.cz>
23676 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23678 (possible_polymorphic_call_targets): Use it.
23679 (ipa_devirt): Likewise.
23681 2018-01-12 Martin Liska <mliska@suse.cz>
23683 * profile-count.h (enum profile_quality): Use 0 as invalid
23684 enum value of profile_quality.
23686 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
23688 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23689 -mext-string options.
23691 2018-01-12 Richard Biener <rguenther@suse.de>
23693 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23694 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23695 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23697 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23699 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23701 * configure.ac (--with-long-double-format): Add support for the
23702 configuration option to change the default long double format on
23704 * config.gcc (powerpc*-linux*-*): Likewise.
23705 * configure: Regenerate.
23706 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23707 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23708 used without modification.
23710 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23712 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23713 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23714 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23715 MISC_BUILTIN_SPEC_BARRIER.
23716 (rs6000_init_builtins): Likewise.
23717 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23719 (speculation_barrier): New define_insn.
23720 * doc/extend.texi: Document __builtin_speculation_barrier.
23722 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23725 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23726 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23727 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23729 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
23730 integral modes instead of "ss" and "sd".
23731 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23732 vectors with 32-bit and 64-bit elements.
23733 (vecdupssescalarmodesuffix): New mode attribute.
23734 (vec_dup<mode>): Use it.
23736 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
23739 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23740 frame if argument is passed on stack.
23742 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23745 * ree.c (combine_reaching_defs): Optimize also
23746 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23747 reg2=any_extend(exp); reg1=reg2;, formatting fix.
23749 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23751 PR middle-end/83189
23752 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23754 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23756 PR middle-end/83718
23757 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23758 after they are computed.
23760 2018-01-11 Bin Cheng <bin.cheng@arm.com>
23762 PR tree-optimization/83695
23763 * gimple-loop-linterchange.cc
23764 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23765 reset cached scev information after interchange.
23766 (pass_linterchange::execute): Remove call to scev_reset_htab.
23768 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23770 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23771 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23772 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23773 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23774 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23775 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23776 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23777 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23778 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23779 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23780 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23781 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23782 (V_lane_reg): Likewise.
23783 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23785 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23786 (vfmal_lane_low<mode>_intrinsic,
23787 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23788 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23789 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23790 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23791 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23792 vfmsl_lane_high<mode>_intrinsic): New define_insns.
23794 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23796 * config/arm/arm-cpus.in (fp16fml): New feature.
23797 (ALL_SIMD): Add fp16fml.
23798 (armv8.2-a): Add fp16fml as an option.
23799 (armv8.3-a): Likewise.
23800 (armv8.4-a): Add fp16fml as part of fp16.
23801 * config/arm/arm.h (TARGET_FP16FML): Define.
23802 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23804 * config/arm/arm-modes.def (V2HF): Define.
23805 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23806 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23807 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23808 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23809 vfmsl_low, vfmsl_high): New set of builtins.
23810 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23811 (vfml_op): New code attribute.
23812 (VFMLHALVES): New int iterator.
23813 (VFML, VFMLSEL): New mode attributes.
23814 (V_reg): Define mapping for V2HF.
23815 (V_hi, V_lo): New mode attributes.
23816 (VF_constraint): Likewise.
23817 (vfml_half, vfml_half_selector): New int attributes.
23818 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23820 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23821 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23823 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23824 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23825 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23826 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23828 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23829 Document new effective target and option set.
23831 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23833 * config/arm/arm-cpus.in (armv8_4): New feature.
23834 (ARMv8_4a): New fgroup.
23835 (armv8.4-a): New arch.
23836 * config/arm/arm-tables.opt: Regenerate.
23837 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23838 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23839 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23840 Add matching rules for -march=armv8.4-a and extensions.
23841 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23843 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23846 * config/rx/rx.md (BW): New mode attribute.
23847 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23849 2018-01-11 Richard Biener <rguenther@suse.de>
23851 PR tree-optimization/83435
23852 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23853 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23854 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23856 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23857 Alan Hayward <alan.hayward@arm.com>
23858 David Sherwood <david.sherwood@arm.com>
23860 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23862 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23863 (aarch64_print_address_internal): Use it to check for a zero offset.
23865 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23866 Alan Hayward <alan.hayward@arm.com>
23867 David Sherwood <david.sherwood@arm.com>
23869 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23870 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23871 Return a poly_int64 rather than a HOST_WIDE_INT.
23872 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23873 rather than a HOST_WIDE_INT.
23874 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23875 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23876 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23877 final_offset from HOST_WIDE_INT to poly_int64.
23878 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23879 to_constant when getting the number of units in an Advanced SIMD
23881 (aarch64_builtin_vectorized_function): Check for a constant number
23883 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23885 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23886 attribute instead of GET_MODE_NUNITS.
23887 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23888 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23889 GET_MODE_SIZE for fixed-size registers.
23890 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23891 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23892 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23893 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23894 (aarch64_print_operand, aarch64_print_address_internal)
23895 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23896 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23897 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23898 Handle polynomial GET_MODE_SIZE.
23899 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23900 wider than SImode without modification.
23901 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23902 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23903 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23904 passing and returning SVE modes.
23905 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23906 rather than GEN_INT.
23907 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23908 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23909 (aarch64_allocate_and_probe_stack_space): Likewise.
23910 (aarch64_layout_frame): Cope with polynomial offsets.
23911 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23912 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23913 polynomial offsets.
23914 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23915 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23916 poly_int64 rather than a HOST_WIDE_INT.
23917 (aarch64_get_separate_components, aarch64_process_components)
23918 (aarch64_expand_prologue, aarch64_expand_epilogue)
23919 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23920 (aarch64_anchor_offset): New function, split out from...
23921 (aarch64_legitimize_address): ...here.
23922 (aarch64_builtin_vectorization_cost): Handle polynomial
23923 TYPE_VECTOR_SUBPARTS.
23924 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23926 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23927 number of elements from the PARALLEL rather than the mode.
23928 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23929 rather than GET_MODE_BITSIZE.
23930 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23931 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23932 (aarch64_expand_vec_perm_const_1): Handle polynomial
23933 d->perm.length () and d->perm elements.
23934 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23935 Apply to_constant to d->perm elements.
23936 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23937 polynomial CONST_VECTOR_NUNITS.
23938 (aarch64_move_pointer): Take amount as a poly_int64 rather
23940 (aarch64_progress_pointer): Avoid temporary variable.
23941 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23942 the mode attribute instead of GET_MODE.
23944 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23945 Alan Hayward <alan.hayward@arm.com>
23946 David Sherwood <david.sherwood@arm.com>
23948 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23949 x exists before using it.
23950 (aarch64_add_constant_internal): Rename to...
23951 (aarch64_add_offset_1): ...this. Replace regnum with separate
23952 src and dest rtxes. Handle the case in which they're different,
23953 including when the offset is zero. Replace scratchreg with an rtx.
23954 Use 2 additions if there is no spare register into which we can
23955 move a 16-bit constant.
23956 (aarch64_add_constant): Delete.
23957 (aarch64_add_offset): Replace reg with separate src and dest
23958 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23959 Use aarch64_add_offset_1.
23960 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23961 an rtx rather than an int. Take the delta as a poly_int64
23962 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23963 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23964 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23965 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23966 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23967 and aarch64_add_sp.
23968 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23969 aarch64_add_constant.
23971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23973 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23974 Use scalar_float_mode.
23976 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23978 * config/aarch64/aarch64-simd.md
23979 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23980 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23981 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23982 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23983 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23984 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23985 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23986 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23987 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23988 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23990 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23993 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23994 targ_options->x_arm_arch_string is non NULL.
23996 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23998 * config/aarch64/aarch64.h
23999 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
24001 2018-01-11 Sudakshina Das <sudi.das@arm.com>
24004 * expmed.c (emit_store_flag_force): Swap if const op0
24005 and change VOIDmode to mode of op0.
24007 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24009 PR rtl-optimization/83761
24010 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
24011 than bytes to mode_for_size.
24013 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24015 PR middle-end/83189
24016 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
24017 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
24020 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24022 PR middle-end/83575
24023 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
24024 when in layout mode.
24025 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
24026 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
24029 2018-01-10 Michael Collison <michael.collison@arm.com>
24031 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
24032 * config/aarch64/aarch64-option-extension.def: Add
24033 AARCH64_OPT_EXTENSION of 'fp16fml'.
24034 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24035 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
24036 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
24037 * config/aarch64/constraints.md (Ui7): New constraint.
24038 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
24039 (VFMLA_SEL_W): Ditto.
24042 (VFMLA16_LOW): New int iterator.
24043 (VFMLA16_HIGH): Ditto.
24044 (UNSPEC_FMLAL): New unspec.
24045 (UNSPEC_FMLSL): Ditto.
24046 (UNSPEC_FMLAL2): Ditto.
24047 (UNSPEC_FMLSL2): Ditto.
24048 (f16mac): New code attribute.
24049 * config/aarch64/aarch64-simd-builtins.def
24050 (aarch64_fmlal_lowv2sf): Ditto.
24051 (aarch64_fmlsl_lowv2sf): Ditto.
24052 (aarch64_fmlalq_lowv4sf): Ditto.
24053 (aarch64_fmlslq_lowv4sf): Ditto.
24054 (aarch64_fmlal_highv2sf): Ditto.
24055 (aarch64_fmlsl_highv2sf): Ditto.
24056 (aarch64_fmlalq_highv4sf): Ditto.
24057 (aarch64_fmlslq_highv4sf): Ditto.
24058 (aarch64_fmlal_lane_lowv2sf): Ditto.
24059 (aarch64_fmlsl_lane_lowv2sf): Ditto.
24060 (aarch64_fmlal_laneq_lowv2sf): Ditto.
24061 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
24062 (aarch64_fmlalq_lane_lowv4sf): Ditto.
24063 (aarch64_fmlsl_lane_lowv4sf): Ditto.
24064 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
24065 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
24066 (aarch64_fmlal_lane_highv2sf): Ditto.
24067 (aarch64_fmlsl_lane_highv2sf): Ditto.
24068 (aarch64_fmlal_laneq_highv2sf): Ditto.
24069 (aarch64_fmlsl_laneq_highv2sf): Ditto.
24070 (aarch64_fmlalq_lane_highv4sf): Ditto.
24071 (aarch64_fmlsl_lane_highv4sf): Ditto.
24072 (aarch64_fmlalq_laneq_highv4sf): Ditto.
24073 (aarch64_fmlsl_laneq_highv4sf): Ditto.
24074 * config/aarch64/aarch64-simd.md:
24075 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
24076 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24077 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
24078 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24079 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
24080 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
24081 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
24082 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
24083 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
24084 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
24085 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
24086 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
24087 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
24088 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
24089 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
24090 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
24091 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
24092 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
24093 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
24094 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
24095 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
24096 (vfmlsl_low_u32): Ditto.
24097 (vfmlalq_low_u32): Ditto.
24098 (vfmlslq_low_u32): Ditto.
24099 (vfmlal_high_u32): Ditto.
24100 (vfmlsl_high_u32): Ditto.
24101 (vfmlalq_high_u32): Ditto.
24102 (vfmlslq_high_u32): Ditto.
24103 (vfmlal_lane_low_u32): Ditto.
24104 (vfmlsl_lane_low_u32): Ditto.
24105 (vfmlal_laneq_low_u32): Ditto.
24106 (vfmlsl_laneq_low_u32): Ditto.
24107 (vfmlalq_lane_low_u32): Ditto.
24108 (vfmlslq_lane_low_u32): Ditto.
24109 (vfmlalq_laneq_low_u32): Ditto.
24110 (vfmlslq_laneq_low_u32): Ditto.
24111 (vfmlal_lane_high_u32): Ditto.
24112 (vfmlsl_lane_high_u32): Ditto.
24113 (vfmlal_laneq_high_u32): Ditto.
24114 (vfmlsl_laneq_high_u32): Ditto.
24115 (vfmlalq_lane_high_u32): Ditto.
24116 (vfmlslq_lane_high_u32): Ditto.
24117 (vfmlalq_laneq_high_u32): Ditto.
24118 (vfmlslq_laneq_high_u32): Ditto.
24119 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
24120 (AARCH64_FL_FOR_ARCH8_4): New.
24121 (AARCH64_ISA_F16FML): New ISA flag.
24122 (TARGET_F16FML): New feature flag for fp16fml.
24123 (doc/invoke.texi): Document new fp16fml option.
24125 2018-01-10 Michael Collison <michael.collison@arm.com>
24127 * config/aarch64/aarch64-builtins.c:
24128 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
24129 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24130 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
24131 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
24132 (AARCH64_ISA_SHA3): New ISA flag.
24133 (TARGET_SHA3): New feature flag for sha3.
24134 * config/aarch64/iterators.md (sha512_op): New int attribute.
24135 (CRYPTO_SHA512): New int iterator.
24136 (UNSPEC_SHA512H): New unspec.
24137 (UNSPEC_SHA512H2): Ditto.
24138 (UNSPEC_SHA512SU0): Ditto.
24139 (UNSPEC_SHA512SU1): Ditto.
24140 * config/aarch64/aarch64-simd-builtins.def
24141 (aarch64_crypto_sha512hqv2di): New builtin.
24142 (aarch64_crypto_sha512h2qv2di): Ditto.
24143 (aarch64_crypto_sha512su0qv2di): Ditto.
24144 (aarch64_crypto_sha512su1qv2di): Ditto.
24145 (aarch64_eor3qv8hi): Ditto.
24146 (aarch64_rax1qv2di): Ditto.
24147 (aarch64_xarqv2di): Ditto.
24148 (aarch64_bcaxqv8hi): Ditto.
24149 * config/aarch64/aarch64-simd.md:
24150 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
24151 (aarch64_crypto_sha512su0qv2di): Ditto.
24152 (aarch64_crypto_sha512su1qv2di): Ditto.
24153 (aarch64_eor3qv8hi): Ditto.
24154 (aarch64_rax1qv2di): Ditto.
24155 (aarch64_xarqv2di): Ditto.
24156 (aarch64_bcaxqv8hi): Ditto.
24157 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
24158 (vsha512h2q_u64): Ditto.
24159 (vsha512su0q_u64): Ditto.
24160 (vsha512su1q_u64): Ditto.
24161 (veor3q_u16): Ditto.
24162 (vrax1q_u64): Ditto.
24163 (vxarq_u64): Ditto.
24164 (vbcaxq_u16): Ditto.
24165 * config/arm/types.md (crypto_sha512): New type attribute.
24166 (crypto_sha3): Ditto.
24167 (doc/invoke.texi): Document new sha3 option.
24169 2018-01-10 Michael Collison <michael.collison@arm.com>
24171 * config/aarch64/aarch64-builtins.c:
24172 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
24173 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24174 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
24175 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
24176 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
24177 (AARCH64_ISA_SM4): New ISA flag.
24178 (TARGET_SM4): New feature flag for sm4.
24179 * config/aarch64/aarch64-simd-builtins.def
24180 (aarch64_sm3ss1qv4si): Ditto.
24181 (aarch64_sm3tt1aq4si): Ditto.
24182 (aarch64_sm3tt1bq4si): Ditto.
24183 (aarch64_sm3tt2aq4si): Ditto.
24184 (aarch64_sm3tt2bq4si): Ditto.
24185 (aarch64_sm3partw1qv4si): Ditto.
24186 (aarch64_sm3partw2qv4si): Ditto.
24187 (aarch64_sm4eqv4si): Ditto.
24188 (aarch64_sm4ekeyqv4si): Ditto.
24189 * config/aarch64/aarch64-simd.md:
24190 (aarch64_sm3ss1qv4si): Ditto.
24191 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
24192 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
24193 (aarch64_sm4eqv4si): Ditto.
24194 (aarch64_sm4ekeyqv4si): Ditto.
24195 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
24196 (sm3part_op): Ditto.
24197 (CRYPTO_SM3TT): Ditto.
24198 (CRYPTO_SM3PART): Ditto.
24199 (UNSPEC_SM3SS1): New unspec.
24200 (UNSPEC_SM3TT1A): Ditto.
24201 (UNSPEC_SM3TT1B): Ditto.
24202 (UNSPEC_SM3TT2A): Ditto.
24203 (UNSPEC_SM3TT2B): Ditto.
24204 (UNSPEC_SM3PARTW1): Ditto.
24205 (UNSPEC_SM3PARTW2): Ditto.
24206 (UNSPEC_SM4E): Ditto.
24207 (UNSPEC_SM4EKEY): Ditto.
24208 * config/aarch64/constraints.md (Ui2): New constraint.
24209 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
24210 * config/arm/types.md (crypto_sm3): New type attribute.
24211 (crypto_sm4): Ditto.
24212 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
24213 (vsm3tt1aq_u32): Ditto.
24214 (vsm3tt1bq_u32): Ditto.
24215 (vsm3tt2aq_u32): Ditto.
24216 (vsm3tt2bq_u32): Ditto.
24217 (vsm3partw1q_u32): Ditto.
24218 (vsm3partw2q_u32): Ditto.
24219 (vsm4eq_u32): Ditto.
24220 (vsm4ekeyq_u32): Ditto.
24221 (doc/invoke.texi): Document new sm4 option.
24223 2018-01-10 Michael Collison <michael.collison@arm.com>
24225 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
24226 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
24227 (AARCH64_FL_FOR_ARCH8_4): New.
24228 (AARCH64_FL_V8_4): New flag.
24229 (doc/invoke.texi): Document new armv8.4-a option.
24231 2018-01-10 Michael Collison <michael.collison@arm.com>
24233 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24234 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
24235 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
24236 * config/aarch64/aarch64-option-extension.def: Add
24237 AARCH64_OPT_EXTENSION of 'sha2'.
24238 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
24239 (crypto): Disable sha2 and aes if crypto disabled.
24240 (crypto): Enable aes and sha2 if enabled.
24241 (simd): Disable sha2 and aes if simd disabled.
24242 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
24244 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
24245 (TARGET_SHA2): New feature flag for sha2.
24246 (TARGET_AES): New feature flag for aes.
24247 * config/aarch64/aarch64-simd.md:
24248 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
24249 conditional on TARGET_AES.
24250 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
24251 (aarch64_crypto_sha1hsi): Make pattern conditional
24253 (aarch64_crypto_sha1hv4si): Ditto.
24254 (aarch64_be_crypto_sha1hv4si): Ditto.
24255 (aarch64_crypto_sha1su1v4si): Ditto.
24256 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
24257 (aarch64_crypto_sha1su0v4si): Ditto.
24258 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
24259 (aarch64_crypto_sha256su0v4si): Ditto.
24260 (aarch64_crypto_sha256su1v4si): Ditto.
24261 (doc/invoke.texi): Document new aes and sha2 options.
24263 2018-01-10 Martin Sebor <msebor@redhat.com>
24265 PR tree-optimization/83781
24266 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
24269 2018-01-11 Martin Sebor <msebor@gmail.com>
24270 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24272 PR tree-optimization/83501
24273 PR tree-optimization/81703
24275 * tree-ssa-strlen.c (get_string_cst): Rename...
24276 (get_string_len): ...to this. Handle global constants.
24277 (handle_char_store): Adjust.
24279 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
24280 Jim Wilson <jimw@sifive.com>
24282 * config/riscv/riscv-protos.h (riscv_output_return): New.
24283 * config/riscv/riscv.c (struct machine_function): New naked_p field.
24284 (riscv_attribute_table, riscv_output_return),
24285 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
24286 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
24287 (riscv_compute_frame_info): Only compute frame->mask if not a naked
24289 (riscv_expand_prologue): Add early return for naked function.
24290 (riscv_expand_epilogue): Likewise.
24291 (riscv_function_ok_for_sibcall): Return false for naked function.
24292 (riscv_set_current_function): New.
24293 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
24294 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
24295 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
24296 * doc/extend.texi (RISC-V Function Attributes): New.
24298 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
24300 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
24301 check for 128-bit long double before checking TCmode.
24302 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
24303 128-bit long doubles before checking TFmode or TCmode.
24304 (FLOAT128_IBM_P): Likewise.
24306 2018-01-10 Martin Sebor <msebor@redhat.com>
24308 PR tree-optimization/83671
24309 * builtins.c (c_strlen): Unconditionally return zero for the empty
24311 Use -Warray-bounds for warnings.
24312 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
24313 for non-constant array indices with COMPONENT_REF, arrays of
24314 arrays, and pointers to arrays.
24315 (gimple_fold_builtin_strlen): Determine and set length range for
24316 non-constant character arrays.
24318 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
24320 PR middle-end/81897
24321 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
24324 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
24326 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
24328 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24331 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
24332 VECTOR_MEM_ALTIVEC_OR_VSX_P.
24333 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
24334 indexed_or_indirect_operand predicate.
24335 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
24336 (*vsx_le_perm_load_v8hi): Likewise.
24337 (*vsx_le_perm_load_v16qi): Likewise.
24338 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
24339 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
24340 (*vsx_le_perm_store_v8hi): Likewise.
24341 (*vsx_le_perm_store_v16qi): Likewise.
24342 (eight unnamed splitters): Likewise.
24344 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24346 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
24347 * config/rs6000/emmintrin.h: Likewise.
24348 * config/rs6000/mmintrin.h: Likewise.
24349 * config/rs6000/xmmintrin.h: Likewise.
24351 2018-01-10 David Malcolm <dmalcolm@redhat.com>
24354 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
24356 * tree.c (tree_nop_conversion): Return true for location wrapper
24358 (maybe_wrap_with_location): New function.
24359 (selftest::check_strip_nops): New function.
24360 (selftest::test_location_wrappers): New function.
24361 (selftest::tree_c_tests): Call it.
24362 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
24363 (maybe_wrap_with_location): New decl.
24364 (EXPR_LOCATION_WRAPPER_P): New macro.
24365 (location_wrapper_p): New inline function.
24366 (tree_strip_any_location_wrapper): New inline function.
24368 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
24371 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
24372 stack_realign_offset for the largest alignment of stack slot
24374 (ix86_find_max_used_stack_alignment): New function.
24375 (ix86_finalize_stack_frame_flags): Use it. Set
24376 max_used_stack_alignment if we don't realign stack.
24377 * config/i386/i386.h (machine_function): Add
24378 max_used_stack_alignment.
24380 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
24382 * config/arm/arm.opt (-mbranch-cost): New option.
24383 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
24386 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
24389 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
24390 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
24392 2018-01-10 Richard Biener <rguenther@suse.de>
24395 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
24396 early out so it also covers the case where we have a non-NULL
24399 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
24401 PR tree-optimization/83753
24402 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
24403 for non-strided grouped accesses if the number of elements is 1.
24405 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24408 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
24409 * i386.h (TARGET_USE_GATHER): Define.
24410 * x86-tune.def (X86_TUNE_USE_GATHER): New.
24412 2018-01-10 Martin Liska <mliska@suse.cz>
24415 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
24416 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
24418 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
24419 CLEANUP_NO_PARTITIONING is not set.
24421 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
24423 * doc/rtl.texi: Remove documentation of (const ...) wrappers
24424 for vectors, as a partial revert of r254296.
24425 * rtl.h (const_vec_p): Delete.
24426 (const_vec_duplicate_p): Don't test for vector CONSTs.
24427 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
24428 * expmed.c (make_tree): Likewise.
24431 * common.md (E, F): Use CONSTANT_P instead of checking for
24433 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
24434 checking for CONST_VECTOR.
24436 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24438 PR middle-end/83575
24439 * predict.c (force_edge_cold): Handle in more sane way edges
24440 with no prediction.
24442 2018-01-09 Carl Love <cel@us.ibm.com>
24444 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24446 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24447 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24448 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24449 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
24450 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24451 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
24452 * config/rs6000/rs6000-protos.h: Add extern defition for
24453 rs6000_generate_float2_double_code.
24454 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24456 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24457 (float2_v2df): Add define_expand.
24459 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
24462 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24463 op_mode in the force_to_mode call.
24465 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24467 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24468 instead of checking each element individually.
24469 (aarch64_evpc_uzp): Likewise.
24470 (aarch64_evpc_zip): Likewise.
24471 (aarch64_evpc_ext): Likewise.
24472 (aarch64_evpc_rev): Likewise.
24473 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24474 instead of checking each element individually. Return true without
24476 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24477 whether all selected elements come from the same input, instead of
24478 checking each element individually. Remove calls to gen_rtx_REG,
24479 start_sequence and end_sequence and instead assert that no rtl is
24482 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24484 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24485 order of HIGH and CONST checks.
24487 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24489 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24490 if the destination isn't an SSA_NAME.
24492 2018-01-09 Richard Biener <rguenther@suse.de>
24494 PR tree-optimization/83668
24495 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24497 (canonicalize_loop_form): ... here, renamed from ...
24498 (canonicalize_loop_closed_ssa_form): ... this and amended to
24499 swap successor edges for loop exit blocks to make us use
24500 the RPO order we need for initial schedule generation.
24502 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24504 PR tree-optimization/64811
24505 * match.pd: When optimizing comparisons with Inf, avoid
24506 introducing or losing exceptions from comparisons with NaN.
24508 2018-01-09 Martin Liska <mliska@suse.cz>
24511 * asan.c (shadow_mem_size): Add gcc_assert.
24513 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24515 Don't save registers in main().
24518 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24519 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24520 * config/avr/avr.c (avr_set_current_function): Don't error if
24521 naked, OS_task or OS_main are specified at the same time.
24522 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24524 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24526 * common/config/avr/avr-common.c (avr_option_optimization_table):
24527 Switch on -mmain-is-OS_task for optimizing compilations.
24529 2018-01-09 Richard Biener <rguenther@suse.de>
24531 PR tree-optimization/83572
24532 * graphite.c: Include cfganal.h.
24533 (graphite_transform_loops): Connect infinite loops to exit
24534 and remove fake edges at the end.
24536 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24538 * ipa-inline.c (edge_badness): Revert accidental checkin.
24540 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24543 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24544 symbols; not inline clones.
24546 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24549 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24550 hard registers. Formatting fixes.
24552 PR preprocessor/83722
24553 * gcc.c (try_generate_repro): Pass
24554 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24555 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24558 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24559 Kito Cheng <kito.cheng@gmail.com>
24561 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24562 (riscv_leaf_function_p): Delete.
24563 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24565 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24567 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24569 (do_ifelse): New function.
24570 (do_isel): New function.
24571 (do_sub3): New function.
24572 (do_add3): New function.
24573 (do_load_mask_compare): New function.
24574 (do_overlap_load_compare): New function.
24575 (expand_compare_loop): New function.
24576 (expand_block_compare): Call expand_compare_loop() when appropriate.
24577 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24578 option description.
24579 (-mblock-compare-inline-loop-limit): New option.
24581 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24584 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24585 Reverse order of second and third operands in first alternative.
24586 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24587 of first and second elements in UNSPEC_VPERMR vector.
24588 (altivec_expand_vec_perm_le): Likewise.
24590 2018-01-08 Jeff Law <law@redhat.com>
24592 PR rtl-optimizatin/81308
24593 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24594 (process_switch): If group_case_labels makes a change, then set
24596 (pass_convert_switch::execute): If a switch is converted, then
24597 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24599 PR rtl-optimization/81308
24600 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24603 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24605 PR target/83663 - Revert r255946
24606 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24607 generation for cases where splatting a value is not useful.
24608 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24609 across a vec_duplicate and a paradoxical subreg forming a vector
24610 mode to a vec_concat.
24612 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24614 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24615 -march=armv8.3-a variants.
24616 * config/arm/t-multilib: Likewise.
24617 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24619 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24621 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24623 (cceq_ior_compare_complement): Give it a name so I can use it, and
24624 change boolean_or_operator predicate to boolean_operator so it can
24625 be used to generate a crand.
24626 (eqne): New code iterator.
24627 (bd/bd_neg): New code_attrs.
24628 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24629 a single define_insn.
24630 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24631 decrement (bdnzt/bdnzf/bdzt/bdzf).
24632 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24633 with the new names of the branch decrement patterns, and added the
24634 names of the branch decrement conditional patterns.
24636 2018-01-08 Richard Biener <rguenther@suse.de>
24638 PR tree-optimization/83563
24639 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24642 2018-01-08 Richard Biener <rguenther@suse.de>
24644 PR middle-end/83713
24645 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24647 2018-01-08 Richard Biener <rguenther@suse.de>
24649 PR tree-optimization/83685
24650 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24651 references to abnormals.
24653 2018-01-08 Richard Biener <rguenther@suse.de>
24656 * dwarf2out.c (output_indirect_strings): Handle empty
24657 skeleton_debug_str_hash.
24658 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24660 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24662 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24663 (emit_store_direct): Likewise.
24664 (arc_trampoline_adjust_address): Likewise.
24665 (arc_asm_trampoline_template): New function.
24666 (arc_initialize_trampoline): Use asm_trampoline_template.
24667 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24668 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24669 * config/arc/arc.md (flush_icache): Delete pattern.
24671 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24673 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24674 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24677 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24680 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24681 by not USED_FOR_TARGET.
24682 (make_pass_resolve_sw_modes): Likewise.
24684 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24686 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24689 2018-01-08 Richard Biener <rguenther@suse.de>
24691 PR middle-end/83580
24692 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24694 2018-01-08 Richard Biener <rguenther@suse.de>
24696 PR middle-end/83517
24697 * match.pd ((t * 2) / 2) -> t): Add missing :c.
24699 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
24701 PR middle-end/81897
24702 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24703 basic blocks with a small number of successors.
24704 (convert_control_dep_chain_into_preds): Improve handling of
24706 (dump_predicates): Split apart into...
24707 (dump_pred_chain): ...here...
24708 (dump_pred_info): ...and here.
24709 (can_one_predicate_be_invalidated_p): Add debugging printfs.
24710 (can_chain_union_be_invalidated_p): Improve check for invalidation
24712 (uninit_uses_cannot_happen): Avoid unnecessary if
24713 convert_control_dep_chain_into_preds yielded nothing.
24715 2018-01-06 Martin Sebor <msebor@redhat.com>
24717 PR tree-optimization/83640
24718 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24719 subtracting negative offset from size.
24720 (builtin_access::overlap): Adjust offset bounds of the access to fall
24721 within the size of the object if possible.
24723 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
24725 PR rtl-optimization/83699
24726 * expmed.c (extract_bit_field_1): Restrict the vector usage of
24727 extract_bit_field_as_subreg to cases in which the extracted
24728 value is also a vector.
24730 * lra-constraints.c (process_alt_operands): Test for the equivalence
24731 substitutions when detecting a possible reload cycle.
24733 2018-01-06 Jakub Jelinek <jakub@redhat.com>
24736 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24737 by default if flag_selective_schedling{,2}. Formatting fixes.
24739 PR rtl-optimization/83682
24740 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24741 if it has non-VECTOR_MODE element mode.
24742 (vec_duplicate_p): Likewise.
24744 PR middle-end/83694
24745 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24746 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24748 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24751 * config/i386/i386-builtin.def
24752 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24753 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24754 Require also OPTION_MASK_ISA_AVX512F in addition to
24755 OPTION_MASK_ISA_GFNI.
24756 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24757 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24758 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24759 to OPTION_MASK_ISA_GFNI.
24760 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24761 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24762 OPTION_MASK_ISA_AVX512BW.
24763 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24764 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24765 addition to OPTION_MASK_ISA_GFNI.
24766 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24767 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24768 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24769 to OPTION_MASK_ISA_GFNI.
24770 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24771 a requirement for all ISAs rather than any of them with a few
24773 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24775 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24776 bitmasks to be enabled with 3 exceptions, instead of requiring any
24777 enabled ISA with lots of exceptions.
24778 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24779 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24780 Change avx512bw in isa attribute to avx512f.
24781 * config/i386/sgxintrin.h: Add license boilerplate.
24782 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
24783 to __AVX512F__ and __AVX512VL to __AVX512VL__.
24784 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24785 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24787 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24788 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24789 temporarily sse2 rather than sse if not enabled already.
24792 * config/i386/sse.md (VI248_VLBW): Rename to ...
24793 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
24794 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24795 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24796 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24797 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24798 mode iterator instead of VI248_VLBW.
24800 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
24802 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24803 (record_modified): Skip clobbers; add debug output.
24804 (param_change_prob): Use sreal frequencies.
24806 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24808 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24809 punt for user-aligned variables.
24811 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24813 * tree-chrec.c (chrec_contains_symbols): Return true for
24816 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24819 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24820 of (x|y) == x for BICS pattern.
24822 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24824 PR tree-optimization/83605
24825 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24826 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24829 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24831 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24832 * config/epiphany/rtems.h: New file.
24834 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24835 Uros Bizjak <ubizjak@gmail.com>
24838 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24839 QIreg_operand instead of register_operand predicate.
24840 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24841 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24842 comments instead of -fmitigate[-_]rop.
24844 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24847 * cgraphunit.c (symbol_table::compile): Switch to text_section
24848 before calling assembly_start debug hook.
24849 * run-rtl-passes.c (run_rtl_passes): Likewise.
24852 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24854 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24855 range_int_cst_p rather than !symbolic_range_p before calling
24856 extract_range_from_multiplicative_op_1.
24858 2018-01-04 Jeff Law <law@redhat.com>
24860 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24861 redundant test in assertion.
24863 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24865 * doc/rtl.texi: Document machine_mode wrapper classes.
24867 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24869 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24870 using tree_to_uhwi.
24872 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24874 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24875 the VEC_PERM_EXPR fold to fail.
24877 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24880 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24881 to switched_sections.
24883 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24886 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24887 test for d.testing.
24889 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24892 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24893 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24895 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24898 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24899 is BLKmode and bitpos not zero or mode change is needed.
24901 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24904 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24907 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24910 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24911 instead of MULT rtx. Update all corresponding splitters.
24912 (*saddl_se): Ditto.
24913 (*ssub<modesuffix>): Ditto.
24914 (*ssubl_se): Ditto.
24915 (*cmp_sadd_di): Update split patterns.
24916 (*cmp_sadd_si): Ditto.
24917 (*cmp_sadd_sidi): Ditto.
24918 (*cmp_ssub_di): Ditto.
24919 (*cmp_ssub_si): Ditto.
24920 (*cmp_ssub_sidi): Ditto.
24921 * config/alpha/predicates.md (const23_operand): New predicate.
24922 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24923 Look for ASHIFT, not MULT inner operand.
24924 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24926 2018-01-04 Martin Liska <mliska@suse.cz>
24928 PR gcov-profile/83669
24929 * gcov.c (output_intermediate_file): Add version to intermediate
24931 * doc/gcov.texi: Document new field 'version' in intermediate
24932 file format. Fix location of '-k' option of gcov command.
24934 2018-01-04 Martin Liska <mliska@suse.cz>
24937 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24939 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24941 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24943 2018-01-03 Martin Sebor <msebor@redhat.com>
24945 PR tree-optimization/83655
24946 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24947 checking calls with invalid arguments.
24949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24951 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24952 (vectorizable_mask_load_store): Delete.
24953 (vectorizable_call): Return false for masked loads and stores.
24954 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24955 instead of gimple_assign_rhs1.
24956 (vectorizable_load): Handle IFN_MASK_LOAD.
24957 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24961 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24963 (vectorizable_mask_load_store): ...here.
24964 (vectorizable_load): ...and here.
24966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24968 * tree-vect-stmts.c (vect_build_all_ones_mask)
24969 (vect_build_zero_merge_argument): New functions, split out from...
24970 (vectorizable_load): ...here.
24972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24974 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24976 (vectorizable_mask_load_store): ...here.
24977 (vectorizable_store): ...and here.
24979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24981 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24983 (vectorizable_mask_load_store): ...here.
24985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24987 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24988 (vect_model_store_cost): Take a vec_load_store_type instead of a
24990 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24991 (vect_model_store_cost): Take a vec_load_store_type instead of a
24993 (vectorizable_mask_load_store): Update accordingly.
24994 (vectorizable_store): Likewise.
24995 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24999 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
25000 IFN_MASK_LOAD calls here rather than...
25001 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
25003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25004 Alan Hayward <alan.hayward@arm.com>
25005 David Sherwood <david.sherwood@arm.com>
25007 * expmed.c (extract_bit_field_1): For vector extracts,
25008 fall back to extract_bit_field_as_subreg if vec_extract
25011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25012 Alan Hayward <alan.hayward@arm.com>
25013 David Sherwood <david.sherwood@arm.com>
25015 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
25016 they are variable or constant sized.
25017 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
25018 slots for constant-sized data.
25020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25021 Alan Hayward <alan.hayward@arm.com>
25022 David Sherwood <david.sherwood@arm.com>
25024 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
25025 handling COND_EXPRs with boolean comparisons, try to find a better
25026 basis for the mask type than the boolean itself.
25028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25030 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
25031 is calculated and how it can be overridden.
25032 * genmodes.c (max_bitsize_mode_any_mode): New variable.
25033 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
25035 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
25038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25039 Alan Hayward <alan.hayward@arm.com>
25040 David Sherwood <david.sherwood@arm.com>
25042 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
25043 Remove the mode argument.
25044 (aarch64_simd_valid_immediate): Remove the mode and inverse
25046 * config/aarch64/iterators.md (bitsize): New iterator.
25047 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
25048 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
25049 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
25050 aarch64_simd_valid_immediate.
25051 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
25052 (aarch64_reg_or_bic_imm): Likewise.
25053 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
25054 with an insn_type enum and msl with a modifier_type enum.
25055 Replace element_width with a scalar_mode. Change the shift
25056 to unsigned int. Add constructors for scalar_float_mode and
25057 scalar_int_mode elements.
25058 (aarch64_vect_float_const_representable_p): Delete.
25059 (aarch64_can_const_movi_rtx_p)
25060 (aarch64_simd_scalar_immediate_valid_for_move)
25061 (aarch64_simd_make_constant): Update call to
25062 aarch64_simd_valid_immediate.
25063 (aarch64_advsimd_valid_immediate_hs): New function.
25064 (aarch64_advsimd_valid_immediate): Likewise.
25065 (aarch64_simd_valid_immediate): Remove mode and inverse
25066 arguments. Rewrite to use the above. Use const_vec_duplicate_p
25067 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
25068 and aarch64_float_const_representable_p on the result.
25069 (aarch64_output_simd_mov_immediate): Remove mode argument.
25070 Update call to aarch64_simd_valid_immediate and use of
25071 simd_immediate_info.
25072 (aarch64_output_scalar_simd_mov_immediate): Update call
25075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25076 Alan Hayward <alan.hayward@arm.com>
25077 David Sherwood <david.sherwood@arm.com>
25079 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
25080 (mode_nunits): Likewise CONST_MODE_NUNITS.
25081 * machmode.def (ADJUST_NUNITS): Document.
25082 * genmodes.c (mode_data::need_nunits_adj): New field.
25083 (blank_mode): Update accordingly.
25084 (adj_nunits): New variable.
25085 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
25087 (emit_mode_size_inline): Set need_bytesize_adj for all modes
25088 listed in adj_nunits.
25089 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
25090 listed in adj_nunits. Don't emit case statements for such modes.
25091 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
25092 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
25093 nothing if adj_nunits is nonnull.
25094 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
25095 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
25096 (emit_mode_fbit): Update use of print_maybe_const_decl.
25097 (emit_move_size): Likewise. Treat the array as non-const
25099 (emit_mode_adjustments): Handle adj_nunits.
25101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25103 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
25104 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
25105 (VECTOR_MODES): Use it.
25106 (make_vector_modes): Take the prefix as an argument.
25108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25109 Alan Hayward <alan.hayward@arm.com>
25110 David Sherwood <david.sherwood@arm.com>
25112 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
25113 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
25114 for MODE_VECTOR_BOOL.
25115 * machmode.def (VECTOR_BOOL_MODE): Document.
25116 * genmodes.c (VECTOR_BOOL_MODE): New macro.
25117 (make_vector_bool_mode): New function.
25118 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
25120 * lto-streamer-in.c (lto_input_mode_table): Likewise.
25121 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
25123 * stor-layout.c (int_mode_for_mode): Likewise.
25124 * tree.c (build_vector_type_for_mode): Likewise.
25125 * varasm.c (output_constant_pool_2): Likewise.
25126 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
25127 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
25128 for MODE_VECTOR_BOOL.
25129 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
25130 of mode class checks.
25131 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
25132 instead of a list of mode class checks.
25133 (expand_vector_scalar_condition): Likewise.
25134 (type_for_widest_vector_mode): Handle BImode as an inner mode.
25136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25137 Alan Hayward <alan.hayward@arm.com>
25138 David Sherwood <david.sherwood@arm.com>
25140 * machmode.h (mode_size): Change from unsigned short to
25142 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
25143 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25144 or if measurement_type is not polynomial.
25145 (fixed_size_mode::includes_p): Check for constant-sized modes.
25146 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
25147 return a poly_uint16 rather than an unsigned short.
25148 (emit_mode_size): Change the type of mode_size from unsigned short
25149 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
25150 (emit_mode_adjustments): Cope with polynomial vector sizes.
25151 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25153 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25155 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
25156 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
25157 * caller-save.c (setup_save_areas): Likewise.
25158 (replace_reg_with_saved_mem): Likewise.
25159 * calls.c (emit_library_call_value_1): Likewise.
25160 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
25161 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
25162 (gen_lowpart_for_combine): Likewise.
25163 * convert.c (convert_to_integer_1): Likewise.
25164 * cse.c (equiv_constant, cse_insn): Likewise.
25165 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
25166 (cselib_subst_to_values): Likewise.
25167 * dce.c (word_dce_process_block): Likewise.
25168 * df-problems.c (df_word_lr_mark_ref): Likewise.
25169 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
25170 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
25171 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
25172 (rtl_for_decl_location): Likewise.
25173 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
25174 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
25175 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
25176 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
25177 (expand_expr_real_1): Likewise.
25178 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
25179 (pad_below): Likewise.
25180 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25181 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
25182 * ira.c (get_subreg_tracking_sizes): Likewise.
25183 * ira-build.c (ira_create_allocno_objects): Likewise.
25184 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
25185 (ira_sort_regnos_for_alter_reg): Likewise.
25186 * ira-costs.c (record_operand_costs): Likewise.
25187 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
25188 (resolve_simple_move): Likewise.
25189 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
25190 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
25191 (lra_constraints): Likewise.
25192 (CONST_POOL_OK_P): Reject variable-sized modes.
25193 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
25194 (add_pseudo_to_slot, lra_spill): Likewise.
25195 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25196 * optabs-query.c (get_best_extraction_insn): Likewise.
25197 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25198 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
25199 (expand_mult_highpart, valid_multiword_target_p): Likewise.
25200 * recog.c (offsettable_address_addr_space_p): Likewise.
25201 * regcprop.c (maybe_mode_change): Likewise.
25202 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
25203 * regrename.c (build_def_use): Likewise.
25204 * regstat.c (dump_reg_info): Likewise.
25205 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
25206 (find_reloads, find_reloads_subreg_address): Likewise.
25207 * reload1.c (eliminate_regs_1): Likewise.
25208 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
25209 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
25210 (simplify_binary_operation_1, simplify_subreg): Likewise.
25211 * targhooks.c (default_function_arg_padding): Likewise.
25212 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
25213 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
25214 (verify_gimple_assign_ternary): Likewise.
25215 * tree-inline.c (estimate_move_cost): Likewise.
25216 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25217 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
25218 (get_address_cost_ainc): Likewise.
25219 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
25220 (vect_supportable_dr_alignment): Likewise.
25221 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25222 (vectorizable_reduction): Likewise.
25223 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
25224 (vectorizable_operation, vectorizable_load): Likewise.
25225 * tree.c (build_same_sized_truth_vector_type): Likewise.
25226 * valtrack.c (cleanup_auto_inc_dec): Likewise.
25227 * var-tracking.c (emit_note_insn_var_location): Likewise.
25228 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
25229 (ADDR_VEC_ALIGN): Likewise.
25231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25232 Alan Hayward <alan.hayward@arm.com>
25233 David Sherwood <david.sherwood@arm.com>
25235 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
25237 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25238 or if measurement_type is polynomial.
25239 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
25240 * combine.c (make_extraction): Likewise.
25241 * dse.c (find_shift_sequence): Likewise.
25242 * dwarf2out.c (mem_loc_descriptor): Likewise.
25243 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
25244 (extract_bit_field, extract_low_bits): Likewise.
25245 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
25246 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
25247 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
25248 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
25249 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25250 * reload.c (find_reloads): Likewise.
25251 * reload1.c (alter_reg): Likewise.
25252 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
25253 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
25254 * tree-if-conv.c (predicate_mem_writes): Likewise.
25255 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
25256 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
25257 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
25258 * valtrack.c (dead_debug_insert_temp): Likewise.
25259 * varasm.c (mergeable_constant_section): Likewise.
25260 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
25262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25263 Alan Hayward <alan.hayward@arm.com>
25264 David Sherwood <david.sherwood@arm.com>
25266 * expr.c (expand_assignment): Cope with polynomial mode sizes
25267 when assigning to a CONCAT.
25269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25270 Alan Hayward <alan.hayward@arm.com>
25271 David Sherwood <david.sherwood@arm.com>
25273 * machmode.h (mode_precision): Change from unsigned short to
25275 (mode_to_precision): Return a poly_uint16 rather than an unsigned
25277 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
25278 or if measurement_type is not polynomial.
25279 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
25280 in which the mode is already known to be a scalar_int_mode.
25281 * genmodes.c (emit_mode_precision): Change the type of mode_precision
25282 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
25284 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25285 for GET_MODE_PRECISION.
25286 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25287 for GET_MODE_PRECISION.
25288 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
25290 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
25291 (expand_field_assignment, make_extraction): Likewise.
25292 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
25293 (get_last_value): Likewise.
25294 * convert.c (convert_to_integer_1): Likewise.
25295 * cse.c (cse_insn): Likewise.
25296 * expr.c (expand_expr_real_1): Likewise.
25297 * lra-constraints.c (simplify_operand_subreg): Likewise.
25298 * optabs-query.c (can_atomic_load_p): Likewise.
25299 * optabs.c (expand_atomic_load): Likewise.
25300 (expand_atomic_store): Likewise.
25301 * ree.c (combine_reaching_defs): Likewise.
25302 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
25303 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
25304 * tree.h (type_has_mode_precision_p): Likewise.
25305 * ubsan.c (instrument_si_overflow): Likewise.
25307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25308 Alan Hayward <alan.hayward@arm.com>
25309 David Sherwood <david.sherwood@arm.com>
25311 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
25312 polynomial numbers of units.
25313 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
25314 (valid_vector_subparts_p): New function.
25315 (build_vector_type): Remove temporary shim and take the number
25316 of units as a poly_uint64 rather than an int.
25317 (build_opaque_vector_type): Take the number of units as a
25318 poly_uint64 rather than an int.
25319 * tree.c (build_vector_from_ctor): Handle polynomial
25320 TYPE_VECTOR_SUBPARTS.
25321 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
25322 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
25323 (build_vector_from_val): If the number of units is variable,
25324 use build_vec_duplicate_cst for constant operands and
25325 VEC_DUPLICATE_EXPR otherwise.
25326 (make_vector_type): Remove temporary is_constant ().
25327 (build_vector_type, build_opaque_vector_type): Take the number of
25328 units as a poly_uint64 rather than an int.
25329 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
25331 * cfgexpand.c (expand_debug_expr): Likewise.
25332 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
25333 (store_constructor, expand_expr_real_1): Likewise.
25334 (const_scalar_mask_from_tree): Likewise.
25335 * fold-const-call.c (fold_const_reduction): Likewise.
25336 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
25337 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
25338 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
25339 (fold_relational_const): Likewise.
25340 (native_interpret_vector): Likewise. Change the size from an
25341 int to an unsigned int.
25342 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
25343 TYPE_VECTOR_SUBPARTS.
25344 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
25345 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
25346 duplicating a non-constant operand into a variable-length vector.
25347 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
25348 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
25349 * ipa-icf.c (sem_variable::equals): Likewise.
25350 * match.pd: Likewise.
25351 * omp-simd-clone.c (simd_clone_subparts): Likewise.
25352 * print-tree.c (print_node): Likewise.
25353 * stor-layout.c (layout_type): Likewise.
25354 * targhooks.c (default_builtin_vectorization_cost): Likewise.
25355 * tree-cfg.c (verify_gimple_comparison): Likewise.
25356 (verify_gimple_assign_binary): Likewise.
25357 (verify_gimple_assign_ternary): Likewise.
25358 (verify_gimple_assign_single): Likewise.
25359 * tree-pretty-print.c (dump_generic_node): Likewise.
25360 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25361 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
25362 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
25363 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
25364 (vect_shift_permute_load_chain): Likewise.
25365 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
25366 (expand_vector_condition, optimize_vector_constructor): Likewise.
25367 (lower_vec_perm, get_compute_type): Likewise.
25368 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25369 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
25370 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
25371 (vect_recog_mask_conversion_pattern): Likewise.
25372 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
25373 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
25374 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25375 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
25376 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
25377 (vectorizable_shift, vectorizable_operation, vectorizable_store)
25378 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
25379 (supportable_widening_operation): Likewise.
25380 (supportable_narrowing_operation): Likewise.
25381 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
25383 * varasm.c (output_constant): Likewise.
25385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25386 Alan Hayward <alan.hayward@arm.com>
25387 David Sherwood <david.sherwood@arm.com>
25389 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
25390 so that both the length == 3 and length != 3 cases set up their
25391 own permute vectors. Add comments explaining why we know the
25392 number of elements is constant.
25393 (vect_permute_load_chain): Likewise.
25395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25396 Alan Hayward <alan.hayward@arm.com>
25397 David Sherwood <david.sherwood@arm.com>
25399 * machmode.h (mode_nunits): Change from unsigned char to
25401 (ONLY_FIXED_SIZE_MODES): New macro.
25402 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
25403 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
25404 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
25406 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
25407 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
25408 or if measurement_type is not polynomial.
25409 * genmodes.c (ZERO_COEFFS): New macro.
25410 (emit_mode_nunits_inline): Make mode_nunits_inline return a
25412 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
25413 Use ZERO_COEFFS when emitting initializers.
25414 * data-streamer.h (bp_pack_poly_value): New function.
25415 (bp_unpack_poly_value): Likewise.
25416 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25417 for GET_MODE_NUNITS.
25418 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25419 for GET_MODE_NUNITS.
25420 * tree.c (make_vector_type): Remove temporary shim and make
25421 the real function take the number of units as a poly_uint64
25422 rather than an int.
25423 (build_vector_type_for_mode): Handle polynomial nunits.
25424 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
25425 * emit-rtl.c (const_vec_series_p_1): Likewise.
25426 (gen_rtx_CONST_VECTOR): Likewise.
25427 * fold-const.c (test_vec_duplicate_folding): Likewise.
25428 * genrecog.c (validate_pattern): Likewise.
25429 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
25430 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25431 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
25432 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
25433 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
25434 * rtlanal.c (subreg_get_info): Likewise.
25435 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25436 (vect_grouped_load_supported): Likewise.
25437 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
25438 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
25439 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
25440 (simplify_const_unary_operation, simplify_binary_operation_1)
25441 (simplify_const_binary_operation, simplify_ternary_operation)
25442 (test_vector_ops_duplicate, test_vector_ops): Likewise.
25443 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25444 instead of CONST_VECTOR_NUNITS.
25445 * varasm.c (output_constant_pool_2): Likewise.
25446 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25447 explicit-encoded elements in the XVEC for variable-length vectors.
25449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25451 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25454 Alan Hayward <alan.hayward@arm.com>
25455 David Sherwood <david.sherwood@arm.com>
25457 * coretypes.h (fixed_size_mode): Declare.
25458 (fixed_size_mode_pod): New typedef.
25459 * builtins.h (target_builtins::x_apply_args_mode)
25460 (target_builtins::x_apply_result_mode): Change type to
25461 fixed_size_mode_pod.
25462 * builtins.c (apply_args_size, apply_result_size, result_vector)
25463 (expand_builtin_apply_args_1, expand_builtin_apply)
25464 (expand_builtin_return): Update accordingly.
25466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25468 * cse.c (hash_rtx_cb): Hash only the encoded elements.
25469 * cselib.c (cselib_hash_rtx): Likewise.
25470 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25471 CONST_VECTOR encoding.
25473 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25474 Jeff Law <law@redhat.com>
25477 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25478 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25479 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25480 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25483 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25484 explicitly probe *sp in a noreturn function if there were any callee
25485 register saves or frame pointer is needed.
25487 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25490 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25491 BLKmode for ternary, binary or unary expressions.
25494 * var-tracking.c (delete_vta_debug_insn): New inline function.
25495 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25496 insns from get_insns () to NULL instead of each bb separately.
25497 Use delete_vta_debug_insn. No longer static.
25498 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25499 delete_vta_debug_insns callers.
25500 * rtl.h (delete_vta_debug_insns): Declare.
25501 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25502 instead of variable_tracking_main.
25504 2018-01-03 Martin Sebor <msebor@redhat.com>
25506 PR tree-optimization/83603
25507 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25508 arguments past the endof the argument list in functions declared
25509 without a prototype.
25510 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25511 Avoid checking when arguments are null.
25513 2018-01-03 Martin Sebor <msebor@redhat.com>
25516 * doc/extend.texi (attribute const): Fix a typo.
25517 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25518 issuing -Wsuggest-attribute for void functions.
25520 2018-01-03 Martin Sebor <msebor@redhat.com>
25522 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25523 offset_int::from instead of wide_int::to_shwi.
25524 (maybe_diag_overlap): Remove assertion.
25525 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25526 * gimple-ssa-sprintf.c (format_directive): Same.
25527 (parse_directive): Same.
25528 (sprintf_dom_walker::compute_format_length): Same.
25529 (try_substitute_return_value): Same.
25531 2018-01-03 Jeff Law <law@redhat.com>
25533 PR middle-end/83654
25534 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25535 non-constant residual for zero at runtime and avoid probing in
25536 that case. Reorganize code for trailing problem to mirror handling
25539 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25541 PR tree-optimization/83501
25542 * tree-ssa-strlen.c (get_string_cst): New.
25543 (handle_char_store): Call get_string_cst.
25545 2018-01-03 Martin Liska <mliska@suse.cz>
25547 PR tree-optimization/83593
25548 * tree-ssa-strlen.c: Include tree-cfg.h.
25549 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25550 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25551 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25553 (strlen_dom_walker::before_dom_children): Call
25554 gimple_purge_dead_eh_edges. Dump tranformation with details
25556 (strlen_dom_walker::before_dom_children): Update call by adding
25557 new argument cleanup_eh.
25558 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25560 2018-01-03 Martin Liska <mliska@suse.cz>
25563 * cif-code.def (VARIADIC_THUNK): New enum value.
25564 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25567 2018-01-03 Jan Beulich <jbeulich@suse.com>
25569 * sse.md (mov<mode>_internal): Tighten condition for when to use
25570 vmovdqu<ssescalarsize> for TI and OI modes.
25572 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25574 Update copyright years.
25576 2018-01-03 Martin Liska <mliska@suse.cz>
25579 * ipa-visibility.c (function_and_variable_visibility): Skip
25580 functions with noipa attribure.
25582 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25584 * gcc.c (process_command): Update copyright notice dates.
25585 * gcov-dump.c (print_version): Ditto.
25586 * gcov.c (print_version): Ditto.
25587 * gcov-tool.c (print_version): Ditto.
25588 * gengtype.c (create_file): Ditto.
25589 * doc/cpp.texi: Bump @copying's copyright year.
25590 * doc/cppinternals.texi: Ditto.
25591 * doc/gcc.texi: Ditto.
25592 * doc/gccint.texi: Ditto.
25593 * doc/gcov.texi: Ditto.
25594 * doc/install.texi: Ditto.
25595 * doc/invoke.texi: Ditto.
25597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25599 * vector-builder.h (vector_builder::m_full_nelts): Change from
25600 unsigned int to poly_uint64.
25601 (vector_builder::full_nelts): Update prototype accordingly.
25602 (vector_builder::new_vector): Likewise.
25603 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25604 (vector_builder::operator ==): Likewise.
25605 (vector_builder::finalize): Likewise.
25606 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25607 Take the number of elements as a poly_uint64 rather than an
25609 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25610 from unsigned int to poly_uint64.
25611 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25612 (vec_perm_indices::new_vector): Likewise.
25613 (vec_perm_indices::length): Likewise.
25614 (vec_perm_indices::nelts_per_input): Likewise.
25615 (vec_perm_indices::input_nelts): Likewise.
25616 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25617 number of elements per input as a poly_uint64 rather than an
25618 unsigned int. Use the original encoding for variable-length
25619 vectors, rather than clamping each individual element.
25620 For the second and subsequent elements in each pattern,
25621 clamp the step and base before clamping their sum.
25622 (vec_perm_indices::series_p): Handle polynomial element counts.
25623 (vec_perm_indices::all_in_range_p): Likewise.
25624 (vec_perm_indices_to_tree): Likewise.
25625 (vec_perm_indices_to_rtx): Likewise.
25626 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25627 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25628 (tree_vector_builder::new_binary_operation): Handle polynomial
25629 element counts. Return false if we need to know the number
25630 of elements at compile time.
25631 * fold-const.c (fold_vec_perm): Punt if the number of elements
25632 isn't known at compile time.
25634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25636 * vec-perm-indices.h (vec_perm_builder): Change element type
25637 from HOST_WIDE_INT to poly_int64.
25638 (vec_perm_indices::element_type): Update accordingly.
25639 (vec_perm_indices::clamp): Handle polynomial element_types.
25640 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25641 (vec_perm_indices::all_in_range_p): Likewise.
25642 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25644 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25645 polynomial vec_perm_indices element types.
25646 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25647 * fold-const.c (fold_vec_perm): Likewise.
25648 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25649 * tree-vect-generic.c (lower_vec_perm): Likewise.
25650 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25651 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25652 element type to HOST_WIDE_INT.
25654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25655 Alan Hayward <alan.hayward@arm.com>
25656 David Sherwood <david.sherwood@arm.com>
25658 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25659 rather than an int. Use plus_constant.
25660 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25661 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25664 Alan Hayward <alan.hayward@arm.com>
25665 David Sherwood <david.sherwood@arm.com>
25667 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25668 a HOST_WIDE_INT to a poly_int64.
25670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25671 Alan Hayward <alan.hayward@arm.com>
25672 David Sherwood <david.sherwood@arm.com>
25674 * calls.c (load_register_parameters): Cope with polynomial
25675 mode sizes. Require a constant size for BLKmode parameters
25676 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25677 forces a parameter to be padded at the lsb end in order to
25678 fill a complete number of words, require the parameter size
25679 to be ordered wrt UNITS_PER_WORD.
25681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25682 Alan Hayward <alan.hayward@arm.com>
25683 David Sherwood <david.sherwood@arm.com>
25685 * reload1.c (spill_stack_slot_width): Change element type
25686 from unsigned int to poly_uint64_pod.
25687 (alter_reg): Treat mode sizes as polynomial.
25689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25690 Alan Hayward <alan.hayward@arm.com>
25691 David Sherwood <david.sherwood@arm.com>
25693 * reload.c (complex_word_subreg_p): New function.
25694 (reload_inner_reg_of_subreg, push_reload): Use it.
25696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25697 Alan Hayward <alan.hayward@arm.com>
25698 David Sherwood <david.sherwood@arm.com>
25700 * lra-constraints.c (process_alt_operands): Reject matched
25701 operands whose sizes aren't ordered.
25702 (match_reload): Refer to this check here.
25704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25705 Alan Hayward <alan.hayward@arm.com>
25706 David Sherwood <david.sherwood@arm.com>
25708 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25709 that the mode size is in the set {1, 2, 4, 8, 16}.
25711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25712 Alan Hayward <alan.hayward@arm.com>
25713 David Sherwood <david.sherwood@arm.com>
25715 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25716 Use plus_constant instead of gen_rtx_PLUS.
25718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25719 Alan Hayward <alan.hayward@arm.com>
25720 David Sherwood <david.sherwood@arm.com>
25722 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25723 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25724 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25725 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25726 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25727 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25728 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25729 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25730 * config/i386/i386.c (ix86_push_rounding): ...this new function.
25731 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25733 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25734 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25735 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25736 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25737 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25738 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25739 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25740 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25741 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25742 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25744 * expr.c (emit_move_resolve_push): Treat the input and result
25745 of PUSH_ROUNDING as a poly_int64.
25746 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25747 (emit_push_insn): Likewise.
25748 * lra-eliminations.c (mark_not_eliminable): Likewise.
25749 * recog.c (push_operand): Likewise.
25750 * reload1.c (elimination_effects): Likewise.
25751 * rtlanal.c (nonzero_bits1): Likewise.
25752 * calls.c (store_one_arg): Likewise. Require the padding to be
25753 known at compile time.
25755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25756 Alan Hayward <alan.hayward@arm.com>
25757 David Sherwood <david.sherwood@arm.com>
25759 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25760 Use plus_constant instead of gen_rtx_PLUS.
25762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25763 Alan Hayward <alan.hayward@arm.com>
25764 David Sherwood <david.sherwood@arm.com>
25766 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25767 rather than an int.
25769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25770 Alan Hayward <alan.hayward@arm.com>
25771 David Sherwood <david.sherwood@arm.com>
25773 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25774 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25775 via stack temporaries. Treat the mode size as polynomial too.
25777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25778 Alan Hayward <alan.hayward@arm.com>
25779 David Sherwood <david.sherwood@arm.com>
25781 * expr.c (expand_expr_real_2): When handling conversions involving
25782 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25783 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
25784 as a poly_uint64 too.
25786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25787 Alan Hayward <alan.hayward@arm.com>
25788 David Sherwood <david.sherwood@arm.com>
25790 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25793 Alan Hayward <alan.hayward@arm.com>
25794 David Sherwood <david.sherwood@arm.com>
25796 * combine.c (can_change_dest_mode): Handle polynomial
25797 REGMODE_NATURAL_SIZE.
25798 * expmed.c (store_bit_field_1): Likewise.
25799 * expr.c (store_constructor): Likewise.
25800 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25801 and polynomial REGMODE_NATURAL_SIZE.
25802 (gen_lowpart_common): Likewise.
25803 * reginfo.c (record_subregs_of_mode): Likewise.
25804 * rtlanal.c (read_modify_subreg_p): Likewise.
25806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25807 Alan Hayward <alan.hayward@arm.com>
25808 David Sherwood <david.sherwood@arm.com>
25810 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25811 numbers of elements.
25813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25814 Alan Hayward <alan.hayward@arm.com>
25815 David Sherwood <david.sherwood@arm.com>
25817 * match.pd: Cope with polynomial numbers of vector elements.
25819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25820 Alan Hayward <alan.hayward@arm.com>
25821 David Sherwood <david.sherwood@arm.com>
25823 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25824 in a POINTER_PLUS_EXPR.
25826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25827 Alan Hayward <alan.hayward@arm.com>
25828 David Sherwood <david.sherwood@arm.com>
25830 * omp-simd-clone.c (simd_clone_subparts): New function.
25831 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25832 (ipa_simd_modify_function_body): Likewise.
25834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25835 Alan Hayward <alan.hayward@arm.com>
25836 David Sherwood <david.sherwood@arm.com>
25838 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25839 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25840 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25841 (expand_vector_condition, vector_element): Likewise.
25842 (subparts_gt): New function.
25843 (get_compute_type): Use subparts_gt.
25844 (count_type_subparts): Delete.
25845 (expand_vector_operations_1): Use subparts_gt instead of
25846 count_type_subparts.
25848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25849 Alan Hayward <alan.hayward@arm.com>
25850 David Sherwood <david.sherwood@arm.com>
25852 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25853 (vect_compile_time_alias): ...this new function. Do the calculation
25854 on poly_ints rather than trees.
25855 (vect_prune_runtime_alias_test_list): Update call accordingly.
25857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25858 Alan Hayward <alan.hayward@arm.com>
25859 David Sherwood <david.sherwood@arm.com>
25861 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25863 (vect_schedule_slp_instance): Likewise.
25865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25866 Alan Hayward <alan.hayward@arm.com>
25867 David Sherwood <david.sherwood@arm.com>
25869 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25870 constant and extern definitions for variable-length vectors.
25871 (vect_get_constant_vectors): Note that the number of units
25872 is known to be constant.
25874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25875 Alan Hayward <alan.hayward@arm.com>
25876 David Sherwood <david.sherwood@arm.com>
25878 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25879 of units as polynomial. Choose between WIDE and NARROW based
25882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25883 Alan Hayward <alan.hayward@arm.com>
25884 David Sherwood <david.sherwood@arm.com>
25886 * tree-vect-stmts.c (simd_clone_subparts): New function.
25887 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25890 Alan Hayward <alan.hayward@arm.com>
25891 David Sherwood <david.sherwood@arm.com>
25893 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25894 vectors as polynomial. Use build_index_vector for
25895 IFN_GOMP_SIMD_LANE.
25897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25898 Alan Hayward <alan.hayward@arm.com>
25899 David Sherwood <david.sherwood@arm.com>
25901 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25902 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25903 for variable-length vectors.
25904 (vectorizable_mask_load_store): Treat the number of units as
25905 polynomial, asserting that it is constant if the condition has
25906 already been enforced.
25907 (vectorizable_store, vectorizable_load): Likewise.
25909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25910 Alan Hayward <alan.hayward@arm.com>
25911 David Sherwood <david.sherwood@arm.com>
25913 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25914 of units as polynomial. Punt if we can't tell at compile time
25915 which vector contains the final result.
25917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25918 Alan Hayward <alan.hayward@arm.com>
25919 David Sherwood <david.sherwood@arm.com>
25921 * tree-vect-loop.c (vectorizable_induction): Treat the number
25922 of units as polynomial. Punt on SLP inductions. Use an integer
25923 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25924 cast of such a series for variable-length floating-point
25927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25928 Alan Hayward <alan.hayward@arm.com>
25929 David Sherwood <david.sherwood@arm.com>
25931 * tree.h (build_index_vector): Declare.
25932 * tree.c (build_index_vector): New function.
25933 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25934 of units as polynomial, forcibly converting it to a constant if
25935 vectorizable_reduction has already enforced the condition.
25936 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25937 to create a {1,2,3,...} vector.
25938 (vectorizable_reduction): Treat the number of units as polynomial.
25939 Choose vectype_in based on the largest scalar element size rather
25940 than the smallest number of units. Enforce the restrictions
25943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25944 Alan Hayward <alan.hayward@arm.com>
25945 David Sherwood <david.sherwood@arm.com>
25947 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25948 number of units as polynomial.
25950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25951 Alan Hayward <alan.hayward@arm.com>
25952 David Sherwood <david.sherwood@arm.com>
25954 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25955 * target.def (autovectorize_vector_sizes): Return the vector sizes
25956 by pointer, using vector_sizes rather than a bitmask.
25957 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25958 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25959 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25961 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25962 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25963 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25964 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25965 * omp-general.c (omp_max_vf): Likewise.
25966 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25967 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25968 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25969 * tree-vect-slp.c (vect_slp_bb): Likewise.
25970 * doc/tm.texi: Regenerate.
25971 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25973 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25974 the vector size as a poly_uint64 rather than an unsigned int.
25975 (current_vector_size): Change from an unsigned int to a poly_uint64.
25976 (get_vectype_for_scalar_type): Update accordingly.
25977 * tree.h (build_truth_vector_type): Take the size and number of
25978 units as a poly_uint64 rather than an unsigned int.
25979 (build_vector_type): Add a temporary overload that takes
25980 the number of units as a poly_uint64 rather than an unsigned int.
25981 * tree.c (make_vector_type): Likewise.
25982 (build_truth_vector_type): Take the number of units as a poly_uint64
25983 rather than an unsigned int.
25985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25986 Alan Hayward <alan.hayward@arm.com>
25987 David Sherwood <david.sherwood@arm.com>
25989 * target.def (get_mask_mode): Take the number of units and length
25990 as poly_uint64s rather than unsigned ints.
25991 * targhooks.h (default_get_mask_mode): Update accordingly.
25992 * targhooks.c (default_get_mask_mode): Likewise.
25993 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25994 * doc/tm.texi: Regenerate.
25996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25997 Alan Hayward <alan.hayward@arm.com>
25998 David Sherwood <david.sherwood@arm.com>
26000 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
26001 * omp-general.c (omp_max_vf): Likewise.
26002 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
26003 (expand_omp_simd): Handle polynomial safelen.
26004 * omp-low.c (omplow_simd_context): Add a default constructor.
26005 (omplow_simd_context::max_vf): Change from int to poly_uint64.
26006 (lower_rec_simd_input_clauses): Update accordingly.
26007 (lower_rec_input_clauses): Likewise.
26009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26010 Alan Hayward <alan.hayward@arm.com>
26011 David Sherwood <david.sherwood@arm.com>
26013 * tree-vectorizer.h (vect_nunits_for_cost): New function.
26014 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
26015 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
26016 (vect_analyze_slp_cost): Likewise.
26017 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
26018 (vect_model_load_cost): Likewise.
26020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26021 Alan Hayward <alan.hayward@arm.com>
26022 David Sherwood <david.sherwood@arm.com>
26024 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
26025 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
26026 from an unsigned int * to a poly_uint64_pod *.
26027 (calculate_unrolling_factor): New function.
26028 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
26030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26031 Alan Hayward <alan.hayward@arm.com>
26032 David Sherwood <david.sherwood@arm.com>
26034 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
26035 from an unsigned int to a poly_uint64.
26036 (_loop_vec_info::slp_unrolling_factor): Likewise.
26037 (_loop_vec_info::vectorization_factor): Change from an int
26039 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
26040 (vect_get_num_vectors): New function.
26041 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
26042 (vect_get_num_copies): Use vect_get_num_vectors.
26043 (vect_analyze_data_ref_dependences): Change max_vf from an int *
26044 to an unsigned int *.
26045 (vect_analyze_data_refs): Change min_vf from an int * to a
26047 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26048 than an unsigned HOST_WIDE_INT.
26049 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
26050 (vect_analyze_data_ref_dependence): Change max_vf from an int *
26051 to an unsigned int *.
26052 (vect_analyze_data_ref_dependences): Likewise.
26053 (vect_compute_data_ref_alignment): Handle polynomial vf.
26054 (vect_enhance_data_refs_alignment): Likewise.
26055 (vect_prune_runtime_alias_test_list): Likewise.
26056 (vect_shift_permute_load_chain): Likewise.
26057 (vect_supportable_dr_alignment): Likewise.
26058 (dependence_distance_ge_vf): Take the vectorization factor as a
26059 poly_uint64 rather than an unsigned HOST_WIDE_INT.
26060 (vect_analyze_data_refs): Change min_vf from an int * to a
26062 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
26063 vfm1 as a poly_uint64 rather than an int. Make the same change
26064 for the returned bound_scalar.
26065 (vect_gen_vector_loop_niters): Handle polynomial vf.
26066 (vect_do_peeling): Likewise. Update call to
26067 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
26068 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
26070 * tree-vect-loop.c (vect_determine_vectorization_factor)
26071 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
26072 (vect_get_known_peeling_cost): Likewise.
26073 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
26074 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
26075 (vect_transform_loop): Likewise. Use the lowest possible VF when
26076 updating the upper bounds of the loop.
26077 (vect_min_worthwhile_factor): Make static. Return an unsigned int
26078 rather than an int.
26079 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
26080 polynomial unroll factors.
26081 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
26082 (vect_make_slp_decision): Likewise.
26083 (vect_supported_load_permutation_p): Likewise, and polynomial
26085 (vect_analyze_slp_cost): Handle polynomial vf.
26086 (vect_slp_analyze_node_operations): Likewise.
26087 (vect_slp_analyze_bb_1): Likewise.
26088 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26089 than an unsigned HOST_WIDE_INT.
26090 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
26091 (vectorizable_load): Handle polynomial vf.
26092 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
26094 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
26096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26097 Alan Hayward <alan.hayward@arm.com>
26098 David Sherwood <david.sherwood@arm.com>
26100 * match.pd: Handle bit operations involving three constants
26101 and try to fold one pair.
26103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26105 * tree-vect-loop-manip.c: Include gimple-fold.h.
26106 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
26107 niters_maybe_zero parameters. Handle other cases besides a step of 1.
26108 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
26109 Add a path that uses a step of VF instead of 1, but disable it
26111 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
26112 and niters_no_overflow parameters. Update calls to
26113 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
26114 Create a new SSA name if the latter choses to use a ste other
26115 than zero, and return it via niters_vector_mult_vf_var.
26116 * tree-vect-loop.c (vect_transform_loop): Update calls to
26117 vect_do_peeling, vect_gen_vector_loop_niters and
26118 slpeel_make_loop_iterate_ntimes.
26119 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
26120 (vect_gen_vector_loop_niters): Update declarations after above changes.
26122 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
26124 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
26125 128-bit round to integer instructions.
26126 (ceil<mode>2): Likewise.
26127 (btrunc<mode>2): Likewise.
26128 (round<mode>2): Likewise.
26130 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26132 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
26133 unaligned VSX load/store on P8/P9.
26134 (expand_block_clear): Allow the use of unaligned VSX
26135 load/store on P8/P9.
26137 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26139 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
26141 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
26142 swap associated with both a load and a store.
26144 2018-01-02 Andrew Waterman <andrew@sifive.com>
26146 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
26147 * config/riscv/riscv.md (clear_cache): Use it.
26149 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
26151 * web.c: Remove out-of-date comment.
26153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26155 * expr.c (fixup_args_size_notes): Check that any existing
26156 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
26157 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
26158 (emit_single_push_insn): ...here.
26160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26162 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
26163 (const_vector_encoded_nelts): New function.
26164 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
26165 (const_vector_int_elt, const_vector_elt): Declare.
26166 * emit-rtl.c (const_vector_int_elt_1): New function.
26167 (const_vector_elt): Likewise.
26168 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
26169 of CONST_VECTOR_ELT.
26171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26173 * expr.c: Include rtx-vector-builder.h.
26174 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
26175 directly on the tree encoding.
26176 (const_vector_from_tree): Likewise.
26177 * optabs.c: Include rtx-vector-builder.h.
26178 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
26179 sequence of "u" values.
26180 * vec-perm-indices.c: Include rtx-vector-builder.h.
26181 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
26182 directly on the vec_perm_indices encoding.
26184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26186 * doc/rtl.texi (const_vector): Describe new encoding scheme.
26187 * Makefile.in (OBJS): Add rtx-vector-builder.o.
26188 * rtx-vector-builder.h: New file.
26189 * rtx-vector-builder.c: Likewise.
26190 * rtl.h (rtx_def::u2): Add a const_vector field.
26191 (CONST_VECTOR_NPATTERNS): New macro.
26192 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
26193 (CONST_VECTOR_DUPLICATE_P): Likewise.
26194 (CONST_VECTOR_STEPPED_P): Likewise.
26195 (CONST_VECTOR_ENCODED_ELT): Likewise.
26196 (const_vec_duplicate_p): Check for a duplicated vector encoding.
26197 (unwrap_const_vec_duplicate): Likewise.
26198 (const_vec_series_p): Check for a non-duplicated vector encoding.
26199 Say that the function only returns true for integer vectors.
26200 * emit-rtl.c: Include rtx-vector-builder.h.
26201 (gen_const_vec_duplicate_1): Delete.
26202 (gen_const_vector): Call gen_const_vec_duplicate instead of
26203 gen_const_vec_duplicate_1.
26204 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
26205 (gen_const_vec_duplicate): Use rtx_vector_builder.
26206 (gen_const_vec_series): Likewise.
26207 (gen_rtx_CONST_VECTOR): Likewise.
26208 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
26209 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26210 Build a new vector rather than modifying a CONST_VECTOR in-place.
26211 (handle_special_swappables): Update call accordingly.
26212 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
26213 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26214 Build a new vector rather than modifying a CONST_VECTOR in-place.
26215 (handle_special_swappables): Update call accordingly.
26217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26219 * simplify-rtx.c (simplify_const_binary_operation): Use
26220 CONST_VECTOR_ELT instead of XVECEXP.
26222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26224 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
26225 the selector elements to be different from the data elements
26226 if the selector is a VECTOR_CST.
26227 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
26228 ssizetype for the selector.
26230 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26232 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
26233 before testing each element individually.
26234 * tree-vect-generic.c (lower_vec_perm): Likewise.
26236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26238 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
26239 * selftest-run-tests.c (selftest::run_tests): Call it.
26240 * vector-builder.h (vector_builder::operator ==): New function.
26241 (vector_builder::operator !=): Likewise.
26242 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
26243 (vec_perm_indices::all_from_input_p): New function.
26244 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26245 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
26246 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
26247 instead of reading the VECTOR_CST directly. Detect whether both
26248 vector inputs are the same before constructing the vec_perm_indices,
26249 and update the number of inputs argument accordingly. Use the
26250 utility functions added above. Only construct sel2 if we need to.
26252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26254 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
26255 the broadcast of the low byte.
26256 (expand_mult_highpart): Use an explicit encoding for the permutes.
26257 * optabs-query.c (can_mult_highpart_p): Likewise.
26258 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
26259 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26260 (vectorizable_bswap): Likewise.
26261 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
26262 explicit encoding for the power-of-2 permutes.
26263 (vect_permute_store_chain): Likewise.
26264 (vect_grouped_load_supported): Likewise.
26265 (vect_permute_load_chain): Likewise.
26267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26269 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
26270 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
26271 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
26272 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26273 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
26274 (vect_gen_perm_mask_any): Likewise.
26276 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26278 * int-vector-builder.h: New file.
26279 * vec-perm-indices.h: Include int-vector-builder.h.
26280 (vec_perm_indices): Redefine as an int_vector_builder.
26281 (auto_vec_perm_indices): Delete.
26282 (vec_perm_builder): Redefine as a stand-alone class.
26283 (vec_perm_indices::vec_perm_indices): New function.
26284 (vec_perm_indices::clamp): Likewise.
26285 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
26286 (vec_perm_indices::new_vector): New function.
26287 (vec_perm_indices::new_expanded_vector): Update for new
26288 vec_perm_indices class.
26289 (vec_perm_indices::rotate_inputs): New function.
26290 (vec_perm_indices::all_in_range_p): Operate directly on the
26291 encoded form, without computing elided elements.
26292 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
26293 encoding. Update for new vec_perm_indices class.
26294 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
26295 the given vec_perm_builder.
26296 (expand_vec_perm_var): Update vec_perm_builder constructor.
26297 (expand_mult_highpart): Use vec_perm_builder instead of
26298 auto_vec_perm_indices.
26299 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
26300 vec_perm_indices instead of auto_vec_perm_indices. Use a single
26301 or double series encoding as appropriate.
26302 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
26303 vec_perm_indices instead of auto_vec_perm_indices.
26304 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26305 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26306 (vect_permute_store_chain): Likewise.
26307 (vect_grouped_load_supported): Likewise.
26308 (vect_permute_load_chain): Likewise.
26309 (vect_shift_permute_load_chain): Likewise.
26310 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26311 (vect_transform_slp_perm_load): Likewise.
26312 (vect_schedule_slp_instance): Likewise.
26313 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26314 (vectorizable_mask_load_store): Likewise.
26315 (vectorizable_bswap): Likewise.
26316 (vectorizable_store): Likewise.
26317 (vectorizable_load): Likewise.
26318 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
26319 vec_perm_indices instead of auto_vec_perm_indices. Use
26320 tree_to_vec_perm_builder to read the vector from a tree.
26321 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
26322 vec_perm_builder instead of a vec_perm_indices.
26323 (have_whole_vector_shift): Use vec_perm_builder and
26324 vec_perm_indices instead of auto_vec_perm_indices. Leave the
26325 truncation to calc_vec_perm_mask_for_shift.
26326 (vect_create_epilog_for_reduction): Likewise.
26327 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
26328 from auto_vec_perm_indices to vec_perm_indices.
26329 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26330 instead of changing individual elements.
26331 (aarch64_vectorize_vec_perm_const): Use new_vector to install
26332 the vector in d.perm.
26333 * config/arm/arm.c (expand_vec_perm_d::perm): Change
26334 from auto_vec_perm_indices to vec_perm_indices.
26335 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26336 instead of changing individual elements.
26337 (arm_vectorize_vec_perm_const): Use new_vector to install
26338 the vector in d.perm.
26339 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
26340 Update vec_perm_builder constructor.
26341 (rs6000_expand_interleave): Likewise.
26342 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
26343 (rs6000_expand_interleave): Likewise.
26345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26347 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
26348 to qimode could truncate the indices.
26349 * optabs.c (expand_vec_perm_var): Likewise.
26351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26353 * Makefile.in (OBJS): Add vec-perm-indices.o.
26354 * vec-perm-indices.h: New file.
26355 * vec-perm-indices.c: Likewise.
26356 * target.h (vec_perm_indices): Replace with a forward class
26358 (auto_vec_perm_indices): Move to vec-perm-indices.h.
26359 * optabs.h: Include vec-perm-indices.h.
26360 (expand_vec_perm): Delete.
26361 (selector_fits_mode_p, expand_vec_perm_var): Declare.
26362 (expand_vec_perm_const): Declare.
26363 * target.def (vec_perm_const_ok): Replace with...
26364 (vec_perm_const): ...this new hook.
26365 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
26366 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
26367 * doc/tm.texi: Regenerate.
26368 * optabs.def (vec_perm_const): Delete.
26369 * doc/md.texi (vec_perm_const): Likewise.
26370 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
26371 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
26372 expand_vec_perm for constant permutation vectors. Assert that
26373 the mode of variable permutation vectors is the integer equivalent
26374 of the mode that is being permuted.
26375 * optabs-query.h (selector_fits_mode_p): Declare.
26376 * optabs-query.c: Include vec-perm-indices.h.
26377 (selector_fits_mode_p): New function.
26378 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
26379 is defined, instead of checking whether the vec_perm_const_optab
26380 exists. Use targetm.vectorize.vec_perm_const instead of
26381 targetm.vectorize.vec_perm_const_ok. Check whether the indices
26382 fit in the vector mode before using a variable permute.
26383 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
26384 vec_perm_indices instead of an rtx.
26385 (expand_vec_perm): Replace with...
26386 (expand_vec_perm_const): ...this new function. Take the selector
26387 as a vec_perm_indices rather than an rtx. Also take the mode of
26388 the selector. Update call to shift_amt_for_vec_perm_mask.
26389 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
26390 Use vec_perm_indices::new_expanded_vector to expand the original
26391 selector into bytes. Check whether the indices fit in the vector
26392 mode before using a variable permute.
26393 (expand_vec_perm_var): Make global.
26394 (expand_mult_highpart): Use expand_vec_perm_const.
26395 * fold-const.c: Includes vec-perm-indices.h.
26396 * tree-ssa-forwprop.c: Likewise.
26397 * tree-vect-data-refs.c: Likewise.
26398 * tree-vect-generic.c: Likewise.
26399 * tree-vect-loop.c: Likewise.
26400 * tree-vect-slp.c: Likewise.
26401 * tree-vect-stmts.c: Likewise.
26402 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
26404 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
26405 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
26406 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
26407 (aarch64_vectorize_vec_perm_const): ...this new function.
26408 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26409 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26410 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
26411 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
26412 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26413 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26414 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
26416 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
26417 check for NEON modes.
26418 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
26419 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
26420 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
26421 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
26423 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
26424 the old VEC_PERM_CONST conditions.
26425 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
26426 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
26427 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
26428 (ia64_vectorize_vec_perm_const_ok): Merge into...
26429 (ia64_vectorize_vec_perm_const): ...this new function.
26430 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
26431 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
26432 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
26433 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
26434 * config/mips/mips.c (mips_expand_vec_perm_const)
26435 (mips_vectorize_vec_perm_const_ok): Merge into...
26436 (mips_vectorize_vec_perm_const): ...this new function.
26437 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
26438 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
26439 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
26440 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
26441 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
26442 (rs6000_expand_vec_perm_const): Delete.
26443 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26445 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26446 (altivec_expand_vec_perm_const_le): Take each operand individually.
26447 Operate on constant selectors rather than rtxes.
26448 (altivec_expand_vec_perm_const): Likewise. Update call to
26449 altivec_expand_vec_perm_const_le.
26450 (rs6000_expand_vec_perm_const): Delete.
26451 (rs6000_vectorize_vec_perm_const_ok): Delete.
26452 (rs6000_vectorize_vec_perm_const): New function.
26453 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26454 an element count and rtx array.
26455 (rs6000_expand_extract_even): Update call accordingly.
26456 (rs6000_expand_interleave): Likewise.
26457 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26458 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26459 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26460 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26461 (rs6000_expand_vec_perm_const): Delete.
26462 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26463 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26464 (altivec_expand_vec_perm_const_le): Take each operand individually.
26465 Operate on constant selectors rather than rtxes.
26466 (altivec_expand_vec_perm_const): Likewise. Update call to
26467 altivec_expand_vec_perm_const_le.
26468 (rs6000_expand_vec_perm_const): Delete.
26469 (rs6000_vectorize_vec_perm_const_ok): Delete.
26470 (rs6000_vectorize_vec_perm_const): New function. Remove stray
26471 reference to the SPE evmerge intructions.
26472 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26473 an element count and rtx array.
26474 (rs6000_expand_extract_even): Update call accordingly.
26475 (rs6000_expand_interleave): Likewise.
26476 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26477 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26479 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26483 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26484 vector mode and that that mode matches the mode of the data
26486 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26487 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26488 directly using expand_vec_perm_1 when forcing selectors into
26490 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26494 * optabs-query.h (can_vec_perm_p): Delete.
26495 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26496 * optabs-query.c (can_vec_perm_p): Split into...
26497 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26498 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26499 particular selector is valid.
26500 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26501 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26502 (vect_grouped_load_supported): Likewise.
26503 (vect_shift_permute_load_chain): Likewise.
26504 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26505 (vect_transform_slp_perm_load): Likewise.
26506 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26507 (vectorizable_bswap): Likewise.
26508 (vect_gen_perm_mask_checked): Likewise.
26509 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26510 implementations of variable permutation vectors into account
26511 when deciding which selector to use.
26512 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26513 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26514 with a false third argument.
26515 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26516 to test whether the constant selector is valid and can_vec_perm_var_p
26517 to test whether a variable selector is valid.
26519 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26521 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26522 * optabs-query.c (can_vec_perm_p): Likewise.
26523 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26524 instead of vec_perm_indices.
26525 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26526 (vect_gen_perm_mask_checked): Likewise,
26527 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26528 (vect_gen_perm_mask_checked): Likewise,
26530 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26532 * optabs-query.h (qimode_for_vec_perm): Declare.
26533 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26534 (qimode_for_vec_perm): ...this new function.
26535 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26537 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26539 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26540 does not have a conditional at the top.
26542 2018-01-02 Richard Biener <rguenther@suse.de>
26544 * ipa-inline.c (big_speedup_p): Fix expression.
26546 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26549 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26552 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26556 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26557 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26558 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26559 cond_taken_branch_cost 3->4.
26561 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26563 PR tree-optimization/83581
26564 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26565 TODO_cleanup_cfg if any changes have been made.
26567 PR middle-end/83608
26568 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26569 convert_modes if target mode has the right side, but different mode
26572 PR middle-end/83609
26573 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26574 last argument when extracting from CONCAT. If either from_real or
26575 from_imag is NULL, use expansion through memory. If result is not
26576 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26577 the parts directly to inner mode, if even that fails, use expansion
26580 PR middle-end/83623
26581 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26582 check for bswap in mode rather than HImode and use that in expand_unop
26585 Copyright (C) 2018 Free Software Foundation, Inc.
26587 Copying and distribution of this file, with or without modification,
26588 are permitted in any medium without royalty provided the copyright
26589 notice and this notice are preserved.